Automatic real-time analysis of human sleep stages by an interval histogram method

Recurrent Deep Neural Networks for Real-Time Sleep Stage Classification From Single Channel EEG

Objective: We investigate the design of deep recurrent neural networks for detecting sleep stages from single channel EEG signals recorded at home by non-expert users. We report the effect of data set size, architecture choices, regularization, and personalization on the classification performance. Methods: We evaluated 58 different architectures and training configurations using three-fold cross validation. Results: A network consisting of convolutional (CONV) layers and long short term memory (LSTM) layers can achieve an agreement with a human annotator of Cohen's Kappa of ~0.73 using a training data set of 19 subjects. Regularization and personalization do not lead to a performance gain. Conclusion: The optimal neural network architecture achieves a performance that is very close to the previously reported human inter-expert agreement of Kappa 0.75. Significance: We give the first detailed account of CONV/LSTM network design process for EEG sleep staging in single channel home based setting.

The EEG as an index of neuromodulator balance in memory and mental illness

There is a strong correlation between signature EEG frequency patterns and the relative levels of distinct neuromodulators. These associations become particularly evident during the sleep-wake cycle. The monoamine-acetylcholine balance hypothesis is a theory of neurophysiological markers of the EEG and a detailed description of the findings that support this proposal are presented in this paper. According to this model alpha rhythm reflects the relative predominance of cholinergic muscarinic signals and delta rhythm that of monoaminergic receptor effects. Both high voltage synchronized rhythms are likely mediated by inhibitory Gαi/o-mediated transduction of inhibitory interneurons. Cognitively, alpha and delta EEG measures are proposed to indicate automatic and flexible strategies, respectively. Sleep is associated with marked changes in relative neuromodulator levels corresponding to EEG markers of distinct stages. Sleep studies on memory consolidation present some of the strongest evidence yet for the respective roles of monoaminergic and cholinergic projections in declarative and non-declarative memory processes, a key theoretical premise for understanding the data. Affective dysregulation is reflected in altered EEG patterns during sleep.

FASTER: an unsupervised fully automated sleep staging method for mice

Identifying the stages of sleep, or sleep staging, is an unavoidable step in sleep research and typically requires visual inspection of electroencephalography (EEG) and electromyography (EMG) data. Currently, scoring is slow, biased and prone to error by humans and thus is the most important bottleneck for large-scale sleep research in animals. We have developed an unsupervised, fully automated sleep staging method for mice that allows less subjective and high-throughput evaluation of sleep. Fully Automated Sleep sTaging method via EEG/EMG Recordings (FASTER) is based on nonparametric density estimation clustering of comprehensive EEG/EMG power spectra. FASTER can accurately identify sleep patterns in mice that have been perturbed by drugs or by genetic modification of a clock gene. The overall accuracy is over 90% in every group. 24-h data are staged by a laptop computer in 10 min, which is faster than an experienced human rater. Dramatically improving the sleep staging process in both quality and throughput FASTER will open the door to quantitative and comprehensive animal sleep research.

A transition-constrained discrete hidden Markov model for automatic sleep staging

BACKGROUND

Approximately one-third of the human lifespan is spent sleeping. To diagnose sleep problems, all-night polysomnographic (PSG) recordings including electroencephalograms (EEGs), electrooculograms (EOGs) and electromyograms (EMGs), are usually acquired from the patient and scored by a well-trained expert according to Rechtschaffen & Kales (R&K) rules. Visual sleep scoring is a time-consuming and subjective process. Therefore, the development of an automatic sleep scoring method is desirable.

METHOD

The EEG, EOG and EMG signals from twenty subjects were measured. In addition to selecting sleep characteristics based on the 1968 R&K rules, features utilized in other research were collected. Thirteen features were utilized including temporal and spectrum analyses of the EEG, EOG and EMG signals, and a total of 158  hours of sleep data were recorded. Ten subjects were used to train the Discrete Hidden Markov Model (DHMM), and the remaining ten were tested by the trained DHMM for recognition. Furthermore, the 2-fold cross validation was performed during this experiment.

RESULTS

Overall agreement between the expert and the results presented is 85.29%. With the exception of S1, the sensitivities of each stage were more than 81%. The most accurate stage was SWS (94.9%), and the least-accurately classified stage was S1 (<34%). In the majority of cases, S1 was classified as Wake (21%), S2 (33%) or REM sleep (12%), consistent with previous studies. However, the total time of S1 in the 20 all-night sleep recordings was less than 4%.

CONCLUSION

The results of the experiments demonstrate that the proposed method significantly enhances the recognition rate when compared with prior studies.

AUTOMATED SLEEP STAGING TECHNIQUE BASED ON THE EMPIRICAL MODE DECOMPOSITION ALGORITHM: A PRELIMINARY STUDY

An automatic sleep staging method is proposed to score wakefulness and three nonrapid eye movement (NREM) stages S1, S2 and S3, based on the Empirical Mode Decomposition (EMD) algorithm. Patients with sleep disorders were tested using this method. Good agreements between manual and automatic staging results were achieved in terms of their Cohen's Kappa value.

Automatic analysis of single-channel sleep EEG: validation in healthy individuals

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.

The effect of placebo administration on the first-night effect in healthy young volunteers

The first-night effect is a well-known phenomenon that is considered to result from a subject's lack of adaptation to the unfamiliar environment of a sleep laboratory and to the technical equipment used for polysomnography. The effect has been explored as a laboratory model for transient insomnia. The main characteristics of this effect are short total sleep time (TST) and rapid eye movement (REM) sleep, a lower sleep efficiency index, and longer REM sleep latency. Previous studies have reported that personality traits (such as trait anxiety) are a potential cause of the first-night effect and that the placebo effect is closely related to the anxiety levels of the subjects. To the best of our knowledge, there are no reports regarding the effects of a placebo on first-night sleep. This omission can be explained by the fact that the polysomnographic recordings obtained during the first night of a study are generally excluded from the analysis in order to avoid the inclusion of the first-night effect. In the present study, 8 male university students were subjected to polysomnographic examinations during drug-free, placebo-administration, and benzodiazepine-administration conditions in order to clarify the placebo effect on sleep during consecutive nights, particularly on the first night. The recordings for each condition were conducted for 4 consecutive nights. A placebo or 5 mg nitrazepam was administered at 2230 h using a double-blind crossover design, while no drug was administered during the drug-free condition. There was a 10-day interval between the examination of each condition. Polysomnographic recording was started at 2300 h and continued until the natural awakening of the subjects on the next morning. Subsequently, the subjects were requested to fill in a rating scale that is used to evaluate the subjective perception of sleep. An increase in stage-2 sleep associated with the first-night effect was observed on the first night during the drug-free and placebo-administration conditions. However, REM sleep reduction associated with the first-night effect was detected on the first night during the drug-free condition; this decrease in REM sleep was counteracted by the placebo during the placebo-administration condition. The nitrazepam, but not the placebo, decreased both slow-wave sleep (SWS) and REM sleep. The values for the tendency to fall asleep, feeling refreshed upon awakening in the morning, and the tension upon awakening in the morning were improved to a greater extent by the placebo and nitrazepam administrations than when no drug was administered. These results demonstrate the possibility that placebo administration may have a hypnotic/anxiolytic effect and may improve transient insomnia without causing SWS and REM sleep reductions.

Processed EEG in natural sleep

Sleep stages are conventionally scored according to recommendations by a committee chaired by Rechtschaffen and Kales in 1968. With these rules normal sleep is divided into rapid eye movement sleep and non-rapid eye movement sleep. Non-rapid eye movement sleep is subdivided into four further stages. With the Rechtschaffen and Kales scoring system the dominant sleep stage at any one time is scored. However, the dynamic structure of sleep is not adequately shown. Furthermore, there is considerable interscorer variability, the scoring is time consuming, tedious and difficult to perform. To overcome these limitations automatic sleep scoring devices using processed EEG technology are developed. These developments are discussed in this chapter.

Inter-rater reliability of sleep cyclic alternating pattern (CAP) scoring and validation of a new computer-assisted CAP scoring method

OBJECTIVE

To assess inter-rater reliability between different scorers, from different qualified sleep research groups, in scoring visually the Cyclic Alternating Pattern (CAP), to evaluate the performances of a new tool for the computer-assisted detection of CAP, and to compare its output with the data from the different scorers.

METHODS

CAP was scored in 11 normal sleep recordings by four different raters, coming from three sleep laboratories. CAP was also scored in the same recordings by means of a new computer-assisted method, implemented in the Hypnolab 1.2 (SWS Soft, Italy) software. Data analysis was performed according to the following steps: (a) the inter-rater reliability of CAP parameters between the four different scorers was carried out by means of the Kendall W coefficient of concordance; (b) the analysis of the agreement between the results of the visual and computer-assisted analysis of CAP parameters was also carried out by means of the Kendall W coefficient; (c) a 'consensus' scoring was obtained, for each recording, from the four scorings provided by the different raters, based on the score of the majority of scorers; (d) the degree of agreement between each scorer and the consensus score and between the computer-assisted analysis and the consensus score was quantified by means of the Cohen's k coefficient; (e) the differences between the number of false positive and false negative detections obtained in the visual and in the computer-assisted analysis were also evaluated by means of the non-parametric Wilcoxon test.

RESULTS

The inter-rater reliability of CAP parameters quantified by the Kendall W coefficient of concordance between the four different scorers was high for all the parameters considered and showed values above 0.9 for total CAP time, CAP time in sleep stage 2 and percentage of A phases in sequence; also CAP rate showed a high value (0.829). The most important global parameters of CAP, including total CAP rate and CAP time, scored by the computer-assisted analysis showed a significant concordance with those obtained by the raters. The agreement between the computer-assisted analysis and the consensus scoring for the assignment of the CAP A phase subtype was not distinguishable from that expected from a human scorer. However, the computer-assisted analysis provided a number of false positives and false negatives significantly higher than that of the visual scoring of CAP.

CONCLUSIONS

CAP scoring shows good inter-rater reliability and might be compared in different laboratories the results of which might also be pooled together; however, caution should always be taken because of the variability which can be expected in the classical sleep staging. The computer-assisted detection of CAP can be used with some supervision and correction in large studies when only general parameters such as CAP rate are considered; more editing is necessary for the correct use of the other results.

SIGNIFICANCE

This article describes the first attempt in the literature to evaluate in a detailed way the inter-rater reliability in scoring CAP parameters of normal sleep and the performances of a human-supervised computerized automatic detection system.

Interrater reliability between scorers from eight European sleep laboratories in subjects with different sleep disorders

Interrater variability of sleep stage scorings is a well-known phenomenon. The SIESTA project offered the opportunity to analyse interrater reliability (IRR) between experienced scorers from eight European sleep laboratories within a large sample of patients with different (sleep) disorders: depression, general anxiety disorder with and without non-organic insomnia, Parkinson's disease, period limb movements in sleep and sleep apnoea. The results were based on 196 recordings from 98 patients (73 males: 52.3 +/- 12.1 years and 25 females: 49.5 +/- 11.9 years) for which two independent expert scorings from two different laboratories were available. Cohen's kappa was used to evaluate the IRR on the basis of epochs and intraclass correlation was used to analyse the agreement on quantitative sleep parameters. The overall level of agreement when five different stages were distinguished was kappa = 0.6816 (76.8%), which in terms of kappa reflects a 'substantial' agreement (Landis and Koch, 1977). For different groups of patients kappa values varied from 0.6138 (Parkinson's disease) to 0.8176 (generalized anxiety disorder). With regard to (sleep) stages, the IRR was highest for rapid eye movement (REM), followed by Wake, slow-wave sleep (SWS), non-rapid eye movement 2 (NREM2) and NREM1. The results of regression analysis showed that age and sex only had a statistically significant effect on kappa when the (sleep) stages are considered separately. For NREM2 and SWS a statistically significant decrease of IRR with age has been observed and the IRR for SWS was lower for males than for females. These variations of IRR most probably reflect changes of the sleep electroencephalography (EEG) with age and gender.

Digital tools in polysomnography

Recent advances in the computing power and storage devices have made computer-based recording of polysomnograms (PSGs) very attractive. Digital PSGs offer the possibility of automating many tedious and time-consuming tasks of identifying sleep related events. Automation not only alleviates these laborious tasks, but also introduces a measure of objectivity in the scoring of various discrete events. In this paper we briefly review some automatic methods that have been previously developed by the authors. Automatic sleep staging methods is emphasized with some illustrative results on inter-scorer variations. We also discuss the leg movement event, respiratory event, sleep spindle and rapid-eye-movement detection methods.

Human alpha oscillations in wakefulness, drowsiness period, and REM sleep: different electroencephalographic phenomena within the alpha band

Cortical oscillations in the range of alpha activity (8-13 Hz) are one of the fundamental electrophysiological phenomena of the human electroencephalogram (EEG). Evidence from quantitative EEG data has shown that their electrophysiological features, cortical generation mechanisms, and therefore, their functional correlates vary along the sleep-wake continuum. Specifically, spectral microstructure and EEG coherence levels between anterior and posterior cortical regions permit to differentiate among alpha activity spontaneously appearing in relaxed wakefulness with eyes closed, drowsiness period, and REM sleep, by reflecting distinct properties of neural networks involved in its cortical generation as well as a different interplay between cortical generators, respectively. Besides, the dissimilar spatiotemporal features of brain electrical microstates within the alpha range reveals a different geometry of active neural structures underlying each alpha variant or, simply, changes in the stability level of neural networks during each brain state. Studies reviewed in this paper support the hypothesis that two different alpha variants occur during human REM sleep: 'background responsive alpha activity', blocked over occipital regions when rapid eye movements are present, and 'REM-alpha bursts', non modulated by the alteration of tonic and phasic periods. Altogether, evidence suggests that electrophysiological features of human cortical oscillations in the alpha frequency range vary across different behavioural states, as well as within state, reflecting different cerebral phenomena with probably dissimilar functional meaning.

Computer-assisted sleep staging

To address the subjectivity in manual scoring of polysomnograms, a computer-assisted sleep staging method is presented in this paper. The method uses the principles of segmentation and self-organization (clustering) based on primitive sleep-related features to find the pseudonatural stages present in the record. Sample epochs of these natural stages are presented to the user, who can classify them according to the Rechtschaffen and Kales (RK) or any other standard. The method then learns from these samples to complete the classification. This step allows the active participation of the operator in order to customize the staging to his/her preferences. The method was developed and tested using 12 records of varying types (normal, abnormal, male, female, varying age groups). Results showed an overall concurrence of 80.6% with manual scoring of 20-s epochs according to RK standard. The greatest amount of errors occurred in the identification of the highly transitional Stage 1, 54% of which was misclassified into neighboring stages 2 or Wake.

Circadian rhythms and sleep have additive effects on respiration in the rat

1. We tested two hypotheses: that respiration and metabolism are subject to circadian modulation in wakefulness, non-rapid-eye-movement (NREM) sleep and rapid-eye-movement (REM) sleep; and that the effects of sleep on breathing vary as a function of time of day. 2. Electroencephalogram (EEG), neck electromyogram (EMG) and abdominal body temperature (T(b)) were measured by telemetry in six male Sprague-Dawley rats. The EEG and EMG were used to identify sleep-wake states. Ventilation (V(I)) and metabolic rate (V(CO2)) were measured by plethysmography. Recordings were made over 24 h (12:12 h light:dark) when rats were in established states of wakefulness, NREM sleep and REM sleep. 3. Statistically significant circadian rhythms were observed in V(I) and V(CO2) in each of the wakefulness, NREM sleep and REM sleep states. Amplitudes and phases of the circadian rhythms were similar across sleep-wake states. 4. The circadian rhythm in V(I) was mediated by a circadian rhythm in respiratory frequency (f(R)). Tidal volume (V(T)) was unaffected by time of day in all three sleep-wake states. 5. The 24 h mean V(I) was significantly greater during wakefulness (363.5 +/- 18.5 ml min(-1)) than during NREM sleep (284.8 +/- 11.1 ml min(-1)) and REM sleep (276.1 +/- 13.9 ml min(-1)). V(CO2) and V(T) each significantly decreased from wakefulness to NREM sleep to REM sleep. f(R) was significantly lower in NREM sleep than in wakefulness and REM sleep. 6. These data confirm that ventilation and metabolism exhibit circadian rhythms during wakefulness, and NREM and REM sleep, and refute the hypothesis that state-related effects on breathing vary as a function of time of day. We conclude that the effects of circadian rhythms and sleep-wake state on respiration and metabolic rate are additive in the rat.

Selected Contribution: Regulation of sleep-wake states in response to intermittent hypoxic stimuli applied only in sleep

Recurrent sleep-related hypoxia occurs in common disorders such as obstructive sleep apnea (OSA). The marked changes in sleep after treatment suggest that stimuli associated with OSA (e.g., intermittent hypoxia) may significantly modulate sleep regulation. However, no studies have investigated the independent effects of intermittent sleep-related hypoxia on sleep regulation and recovery sleep after removal of intermittent hypoxia. Ten rats were implanted with telemetry units to record the electroencephalogram (EEG), neck electromyogram, and body temperature. After >7 days recovery, a computer algorithm detected sleep-wake states and triggered hypoxic stimuli (10% O2) or room air stimuli only during sleep for a 3-h period. Sleep-wake states were also recorded for a 3-h recovery period after the stimuli. Each rat received an average of 69.0 +/- 6.9 hypoxic stimuli during sleep. The non-rapid eye movement (non-REM) and rapid-eye-movement (REM) sleep episodes averaged 50.1 +/- 3.2 and 58.9 +/- 6.6 s, respectively, with the hypoxic stimuli, with 32.3 +/- 3.2 and 58.6 +/- 4.8 s of these periods being spent in hypoxia. Compared with results for room air controls, hypoxic stimuli led to increased wakefulness (P < 0.005), nonsignificant changes in non-REM sleep, and reduced REM sleep (P < 0.001). With hypoxic stimuli, wakefulness episodes were longer and more frequent, non-REM periods were shorter and more frequent, and REM episodes were shorter and less frequent (P < 0.015). Hypoxic stimuli also increased faster frequencies in the EEG (P < 0.005). These effects of hypoxic stimuli were reversed on return to room air. There was a rebound increase in REM sleep, increased slower non-REM EEG frequencies, and decreased wakefulness (P < 0.001). The results show that sleep-specific hypoxia leads to significant modulation of sleep-wake regulation both during and after application of the intermittent hypoxic stimuli. This study is the first to determine the independent effects of sleep-related hypoxia on sleep regulation that approximates OSA before and after treatment.

On-line detection of sleep-wake states and application to produce intermittent hypoxia only in sleep in rats

Sleep-disordered breathing is associated with adverse clinical consequences such as daytime sleepiness and hypertension. The mechanisms behind these associations have been studied in animal models, especially rats, but intermittent stimuli such as hypoxia have been applied without reference to sleep-wake states. To determine mechanisms underlying the adverse physiological consequences of stimuli associated with sleep-disordered breathing requires criteria for detection of sleep-wake states on-line to trigger stimuli only in sleep. This study aimed to develop such a system for freely behaving rats. Twelve rats with implanted electroencephalogram and neck electromyogram electrodes were studied in the light and dark phases. Electroencephalogram frequencies in the high (20–30 Hz) and low (2–4 Hz) frequency bands distinguished non-rapid eye movement (REM) sleep, whereas neck electromyogram distinguished REM. Using these parameters in a simple algorithm led to detection accuracies of 94.5 ± 1.0 (SE) % for wakefulness, 96.2 ± 0.8% for non-REM sleep, and 92.3 ± 1.6% for REM compared with blinded human judgment. The algorithm was then used to trigger hypoxic stimuli only in sleep. Because frequency and amplitude analysis is readily performed using a variety of commercial systems, incorporation of these parameters into such an algorithm will facilitate studies investigating mechanisms underlying the physiological consequences of sleep-related respiratory stimuli in a fashion that more effectively models clinical disorders.

Examination of accuracy of sleep stages by means of an automatic sleep analysis system 'Sleep Ukiha'

We examined the differences between the results of an automatic sleep analysis system and inspection decision. Subjects were 10 males (average age 21.6 years). One section consists of 20 s records. The sections that deviated from the algorithm could not be decided. Each sleep stage decided by automatic analysis was compared with the inspection decision. The agreement ratio of stage 3 was 91.6% in the highest, and followed by stage 2, stage 4, stage W and stage 1. The lowest was 62.5% for movement time. The total agreement ratio was 85.8%. The agreement ratios of the apnea index (AI) and the apnea hypopnea index (AHI) were relatively high, but for types of sleep apnea, agreement ratios require improvement.

The effects of diazepam on sleep spindles: a qualitative and quantitative analysis

The effects of 2.5 or 5 mg diazepam (DIZ) on the sleep spindles were studied in 12 healthy male subjects. Polygraphic recordings and the state anxiety scale of the State Trait Anxiety Inventory (STAI) were made for 6 consecutive nights. An inert placebo was given on the first 3 nights and on the sixth night, and DIZ was administered on the fourth and fifth nights. DIZ produced increases in the amount of the slow and fast spindles in a dose-dependent manner. DIZ dose-dependently lowered the frequency of the fast spindles and elevated that of the slow spindles. Furthermore, the influence of DIZ on fast spindles was greater than that on slow spindles. DIZ decreased the state anxiety of the subjects in a dose-dependent manner. These results suggest that measuring the amount and the frequency of fast spindles could be a useful tool in predicting the anxiolytic effects of benzodiazepines.

Alpha sleep characteristics in fibromyalgia

OBJECTIVE

To characterize the patterns of alpha electroencephalographic sleep and their associations with pain and sleep in patients with fibromyalgia.

METHODS

Pain and sleep symptoms of 40 female patients with fibromyalgia and 43 healthy control subjects were studied before and after overnight polysomnography. Blinded analyses of alpha activity in non-rapid eye movement (non-REM) sleep were performed using time domain, frequency domain, and visual analysis techniques.

RESULTS

Three distinct patterns of alpha sleep activity were detected in fibromyalgia: phasic alpha (simultaneous with delta activity) in 50% of patients, tonic alpha (continuous throughout non-REM sleep) in 20% of patients, and low alpha activity in the remaining 30% of patients. Low alpha activity was exhibited by 83.7% of control subjects (P < 0.01). All fibromyalgia patients who displayed phasic alpha sleep, activity reported worsening of pain after sleep, compared with 58.3% of patients with low alpha activity (P < 0.01) and 25.0% of patients with tonic alpha activity (P < 0.01). Postsleep increase in the number of tender points occurred in 90.0% of patients with phasic alpha activity, 41.7% of patients with low alpha activity, and 25.0% of patients with tonic alpha activity (P < 0.01). Self ratings of poor sleep were reported by all patients with phasic alpha activity, 58.3% of patients with low alpha activity (P < 0.01), and 12.5% of patients with tonic alpha activity (P < 0.01). Patients with phasic alpha activity reported longer duration of pain than patients in other subgroups (P < 0.01). Additionally, patients with phasic alpha sleep activity exhibited less total sleep time than patients in other subgroups (P < 0.05), as well as lower sleep efficiency (P < 0.05) and less slow wave sleep (P < 0.05) than patients with a tonic alpha sleep pattern.

CONCLUSION

Alpha intrusion during sleep can be of different patterns. Phasic alpha sleep activity was the pattern that correlated better with clinical manifestations of fibromyalgia.

Slow (0.7-2 Hz) and fast (2-4 Hz) delta components are differently correlated to theta, alpha and beta frequency bands during NREM sleep

OBJECTIVES

Spectral power of 40 all-night sleep EEGs (Cz-Pz bipolar lead) recorded in 20 healthy young subjects was calculated after normalization on 30-s consecutive epochs by means of an autocorrelation method based on a 15-order autoregressive model.

METHODS

The spectral parameters were calculated for the 7 main EEG bands: slow delta (0.7-2 Hz); fast delta (2-4 Hz); theta (4-8 Hz); alpha (8-12 Hz); sigma (12-16 Hz); beta1 (16-35 Hz); and beta 2 (>35 Hz).

RESULTS

Strong negative correlations were found between power in the fast delta and either the alpha or the beta bands and between slow delta and theta bands, whereas the two delta bands showed little correlation with each other.

CONCLUSION

The possibility that theses different relationships of slow and fast delta components with other frequency bands might reflect the neocortical or the thalamocortical origin of the delta waves is discussed.

Automated sleep stage scoring using hybrid rule- and case-based reasoning

We propose an automated method for sleep stage scoring using hybrid rule- and case-based reasoning. The system first performs rule-based sleep stage scoring, according to the Rechtschaffen and Kale's sleep-scoring rule (1968), and then supplements the scoring with case-based reasoning. This method comprises signal processing unit, rule-based scoring unit, and case-based scoring unit. We applied this methodology to three recordings of normal sleep and three recordings of obstructive sleep apnea (OSA). Average agreement rate in normal recordings was 87.5% and case-based scoring enhanced the agreement rate by 5.6%. This architecture showed several advantages over the other analytical approaches in sleep scoring: high performance on sleep disordered recordings, the explanation facility, and the learning ability. The results suggest that combination of rule-based reasoning and case-based reasoning is promising for an automated sleep scoring and it is also considered to be a good model of the cognitive scoring process.

Spectral and Topographic Microstructure of Brain Alpha Activity During Drowsiness at Sleep Onset and REM Sleep

Abstract High-resolution frequency methods were used to describe the spectral and topographic microstructure of human spontaneous alpha activity in the drowsiness (DR) period at sleep onset and during REM sleep. Electroencephalographic (EEG), electrooculographic (EOG), and electromyographic (EMG) measurements were obtained during sleep in 10 healthy volunteer subjects. Spectral microstructure of alpha activity during DR showed a significant maximum power with respect to REM-alpha bursts for the components in the 9.7-10.9 Hz range, whereas REM-alpha bursts reached their maximum statistical differentiation from the sleep onset alpha activity at the components between 7.8 and 8.6 Hz. Furthermore, the maximum energy over occipital regions appeared in a different spectral component in each brain activation state, namely, 10.1 Hz in drowsiness and 8.6 Hz in REM sleep. These results provide quantitative information for differentiating the drowsiness alpha activity and REM-alpha by studying their microstructural properties. On the other hand, these data suggest that the spectral microstructure of alpha activity during sleep onset and REM sleep could be a useful index to implement in automatic classification algorithms in order to improve the differentiation between the two brain states.

Development and application of the 'Sleep Ukiha' automated sleep analysis system

We attempted to develop an automated sleep analysis system that uses a personal computer as an aid to the entire sleep research process. Analysis is based on electroencephalogram, chin muscle electromyogram and electrooculography, while other physiological phenomena can be handled arbitrarily. Major characteristics of the system include: (i) simultaneous gathering of physiological phenomena from up to three patients; (ii) high-speed waveform analysis; (iii) user-friendly operating environment through the use of a graphical user interface; and (iv) versatile utilization of analytical data in research.

Alpha burst activity during human REM sleep: descriptive study and functional hypotheses

OBJECTIVES

Spectral power contribution in the range of alpha activity is a well-known electrophysiological feature of human REM sleep, which could be caused by the spontaneous bursts of alpha activity not associated with arousals that usually appear during this brain state. The present study was undertaken to determine the density of alpha bursts during tonic and phasic (oculomotor) REM periods for each REM cycle. In addition, this phasic brain event was also described from a spectral and topographical point of view.

METHODS

Ten healthy right-handed subjects (5 females) aged 19-25 years (mean 22.9 years, SD 2.6) participated in the present study. Each selected subject filled in a daily sleep log for 2 weeks before the experimental night to provide information on all 3 salient aspects of sleep pattern, sleep experience and sleep effects.

RESULTS

The results revealed that transient REM-alpha bursts, which lasted about 3 s and were accompanied by no increase in the EMG amplitude, appeared in all subjects who participated in this study, showing a higher density in the third and fourth REM cycle during phasic in comparison with tonic periods. The bandpass filtered signals showed the highest spectral contribution for the slower alpha components (8-9 Hz), the occipital scalp regions being the main generator source of this brain activity.

CONCLUSIONS

The authors hypothesize that REM-alpha bursts may work as micro-arousals (or incomplete arousals) facilitating the brain connection with the external world in this cerebral state, whereas REM-alpha arousals - usually longer and accompanied by changes in the EMG amplitude - generate a shift of brain state associated with sleep fragmentation (complete arousal).

Appearance of frontal midline theta activity in patients with generalized anxiety disorder

The appearance of frontal midline theta activity (Fm&theta;), recognized as distinct EEG theta rhythm in the frontal midline area during performance of a mental task, reflects feelings of relief from anxiety in humans. In the present study, EEGs were recorded, and the Hamilton Rating Scale for Anxiety and the state anxiety scale of Spielberger's State-Trait Anxiety Inventory were evaluated once a week in 28 patients with generalized anxiety disorder. The Taylor Manifest Anxiety Scale and the trait anxiety scale of Spielberger's State-Trait Anxiety Inventory were used to assess anxiety before and after the tests. The present results suggest that the appearance of Fm&theta; might be closely related to an improvement in the anxiety symptoms associated with generalized anxiety disorder.

A polysomnographic study of sleep patterns in normal humans with low- or high-anxiety personality traits

To clarify the effects of anxiety-related personality traits on sleep patterns, polysomnographic examinations (PSG) were performed over 4 consecutive nights on normal humans who tested within the low- or high-anxiety ranges. The subjects consisted of two groups of six male university students who scored either less than 45 points (low-anxiety group) or more than 55 points (high-anxiety group) on the Spielberger's State Trait Anxiety Inventory. Compared to the levels of sleep change in the high-anxiety group, the low-anxiety group exhibited a greater change in REM sleep and stage 2 sleep. The REM sleep in the low-anxiety group was shorter on the first and second nights compared to the third and fourth nights, and the stage 2 sleep was longer on the first night than on the remaining three nights. Thus, the low-anxiety group showed a first-night effect followed by partial recovery on the second night, while the high-anxiety group exhibited no obvious first-night effect. These results suggest that there is a difference in sleep patterns, assessed by consecutive PSG, between those with low- and high-anxiety traits, and that anxiety-related personality traits attenuate the occurrence of the first-night effect, reflecting a lower adaptability to a novel environment.

Contribution to the definition of the power bands limits of sleep EEG by linear prediction

The linear prediction method has been applied to compute the power spectra distribution in 10 healthy young men (aged 17 to 26). For each subject, the local extrema positions have been detected and grouped as a function of their frequency. The computation of a mean frequency distribution for the extrema position, on the 10 subjects, allowed for the definition of eight frequency power bands. The time evolution of these power bands during the first NREM episode has been computed and averaged for the 10 subjects. The cross-correlation analysis demonstrated the relevance of three bands in the range 0-8.10 Hz: delta ]0-2.78] Hz, theta 1 ]2.78-5.42] Hz, theta 2 ]5.42-8.10] Hz. Our extrema study leads to the division of the beta range ]16.25-25.00] Hz into three bands. However, their time evolution does not support the usefulness of the partitioning of the beta range, probably resulting from our too low sampling frequency.

Anxiolytic effects of low-dose clomipramine in highly anxious healthy volunteers assessed by frontal midline theta activity

1. The appearance of Fm theta, the distinct EEG theta rhythm in the frontal midline area during performance of a mental task, reflects relief from anxiety in humans. 2. In the present study, the anxiolytic effects of low-dose clomipramine were examined by monitoring the Fm theta amount, the STAI scores and the plasma 5-HIAA concentration in 24 male university students with (Fm theta group, n = 12) and without (non-Fm theta group, n = 12) Fm theta. 3. Subjects were given placebo, 10 mg and 30 mg clomipramine in a double-blind crossover design. Blood samples were obtained, STAI scores were determined, and EEGs were recorded before and during the performance of an arithmetic addition task. The test was repeated twice: before and 3 hrs after drug administration. 4. In the non-Fm theta group, 10 mg clomipramine decreased the 5-HIAA concentration and state anxiety scores but increased the Fm theta amount, while 30 mg clomipramine slightly increased only the Fm theta amount. However, there were no differences in these items before and after clomipramine administration in the Fm theta group. 5. These results suggest that low doses of clomipramine such as 10 mg may exert anxiolytic effects during the acute phase of treatment in highly anxious humans.

FFT analysis of EEG during stage 2-to-REM transitions in narcoleptic patients and normal sleepers

The EEG in REM sleep is markedly different from the EEG in non-REM sleep; however, very little research has been conducted analyzing the transition from NREM to REM sleep. The purpose of this study was to describe the changes in EEG power spectra that coincide with significant physiological events throughout the transition from stage 2 to REM sleep. Furthermore, a comparison of the stage 2 to REM transition between narcoleptic patients and normal sleepers was conducted. Five female and five male patients diagnosed with narcolepsy-cataplexy and 10 normal sleepers matched for age and gender participated. Analyses were based on the consecutive appearance of 3 common physiological indicators of REM sleep: First, a decrease in tonic submental EMG activity; second, the appearance of saw tooth waves; and third, the appearance of rapid eye movements. Systematic changes in EEG power density were evident for delta, theta, alpha and sigma frequencies, across the stage 2-REM transition. These changes appeared to be relatively continuous throughout this transition rather than changing dramatically following the onset of any one of the 3 primary physiological indicators of REM sleep. Furthermore, the transition from stage 2 to stage REM appeared to be similar for narcoleptic patients and normal sleepers.

Differential effects of dopaminergic drugs on anxiety and arousal in healthy volunteers with high and low anxiety

1. The appearance of frontal midline theta activity (Fm theta), the distinct EEG theta rhythm in the frontal midline area during performance of a mental task, indicates relief from anxiety in humans. 2. The authors examined the effects of bromocriptine and sulpiride on anxiety and arousal in 24 male university students with (Fm theta group, n = 12) and without (non-Fm theta group, n = 12) Fm theta. Subjects were given placebo, 2.5 mg bromocriptine and 100 mg sulpiride in a double-blind crossover design. 3. Blood samples were obtained, STAI scores were determined, and EEGs were recorded before and during the performance of an arithmetic addition task. The test was repeated twice: before and 1 hr after drug administration. 4. Bromocriptine reduced the HVA concentration in both groups; sulpiride caused an increase in both groups. In the Fm theta group, bromocriptine did not alter the appearance time of Fm theta, the state anxiety score or the task performance; sulpiride increased the Fm theta amount and reduced the state anxiety but did not affect the task performance. In the non-Fm theta group, bromocriptine increased the Fm theta duration and reduced the state anxiety score but did not influence the task performance, while sulpiride reduced Fm theta and increased the state anxiety but had no effect on the task performance. 5. These results suggest that the sensitivity of presynaptic D2 receptors is higher in high-anxiety subjects compared with low-anxiety subjects, and that anxiolytic effects in high-anxiety humans and those in low-anxiety humans may be caused by decreased and increased DA activity, respectively. In addition, the stimulation of DA function may cause anxiogenic effects in high-anxiety individuals.

Relationship between the time courses of power in the frequency bands of human sleep EEG

The time course of the different frequency bands in the human sleep EEG spectrum within separate NREM and REM episodes averaged over 24 healthy subjects is measured and plotted with the aim of studying their inter-relationship and also the particularities of the beta band. In NREM, a negative sigma-delta cross-correlation corresponding to that expected from neurophysiological data, is found only in the central zone of each episode. The overall correlation is found to be negligible. A neurophysiological explanation is proposed to account for some aspects of this sigma-delta relation. Below the beta frequency range in NREM, the sequence of build-up in power for the different frequencies is from fast to slow, suggesting that there may be a smooth progression in the frequency of oscillation of the thalamocortical neurons depending on their degree of polarization. The beta band is the only one showing a reciprocal relationship with delta throughout the NREM episode, and it is the only one not declining in the REM episode. These differences, together with the close similarity of the beta evolution with that of the REM-on neuronal activity, suggest that beta could directly reflect this activity.

Differential effects of noradrenergic drugs on anxiety and arousal in healthy volunteers with high and low anxiety

1. The appearance of frontal midline theta activity (Fm theta), the distinct EEG theta rhythm in the frontal midline area during performance of a mental task, indicates relief from anxiety in humans. 2. The authors investigated the effects of clonidine and yohimbine on anxiety and arousal in 24 male university students with (Fm theta group, n = 12) and without (non-Fm theta group, n = 12) Fm theta. Subjects received placebo, 0.15 mg clonidine and 15 mg yohimbine in a double-blind crossover design. 3. Blood samples were obtained, state-trait anxiety inventory (STAI) scores were determined, and EEGs were recorded before and during the performance of an arithmetic addition task. The test was repeated twice: before and 1 hr after drug administration. 4. Clonidine reduced the 3-methoxy-4-hydroxyphenylglycol (MHPG) concentration in both groups; yohimbine caused an increase in both groups. In the Fm theta group, clonidine reduced the appearance time of Fm theta and the number of task performance but did not alter the state anxiety scores; yohimbine had no effects on Fm theta or the state anxiety but increased the task performance. In the non-Fm theta group, clonidine increased the Fm theta amount and reduced the state anxiety score but did not affect task performance, while yohimbine reduced Fm theta but increased the state anxiety, the task performance and the number of errors. 5. These results suggest that changes in noradrenaline (NA) activity affect both anxiety and arousal levels in high-anxiety humans, but predominantly affect only the arousal level in low-anxiety humans.

Past and future of computer-assisted sleep analysis and drowsiness assessment

The development of computerized sleep analysis has been very much technology-driven by both mathematical tools and available hardware but, additionally and unfortunately, by the almost-30-year-old standard used for manual sleep stage scoring of paper recordings. There are no technical restrictions in terms of computing power, storage space, and costs anymore. However, the standards of visual sleep stage scoring have proven insufficient and ambiguous, and their utilization evidently provides misleading and erroneous information. The low temporal resolution provided by the one-page epoch, the crude division of the sleep processes into a few discrete stages, and the total ignorance of spatial information are the major drawbacks. It is meaningless to try to improve the computerised systems if the algorithms are based on erroneous concepts. Instead, the focus should be changed to studies dealing with the identification and modelling of true biological sleep-related processes. This work cannot be performed without the successful application of computerized methods, some of which have been used in related fields but have not yet been applied to sleep studies. It is extremely important that basic findings are confirmed with a wide variety of methods in several laboratories. The use of predetermined, fixed criteria for methods, waveforms, and states too early is scientifically erroneous and hazardous. Instead standards should describe the minimum requirements for the recording and analysis of the signals in terms of sampling rate, dynamic range, linearity, and documentation of the methods used. With the development of better technology, these standards ought to be constantly reevaluated and modified. The development toward more open commercial digital systems, including standardized programming methods and data formats, would have great positive impact to the field. These trends have long been established in many other fields of industry.

Stimulatory effect of butoctamide hydrogen succinate on REM sleep in normal humans

1. The efficacy of butoctamide hydrogen succinate (BAHS) was compared with that of nitrazepam on the basis of the polysomnograms and the subjective assessments. 2. Twelve healthy male students were divided into three groups consisting of 4 subjects each with were administered BAHS 600 mg, nitrazepam 5 mg, and BAHS 600 mg + nitrazepam 5 mg, respectively. 3. Polygraphic recordings were made for 8 consecutive nights for each subject, and the polysomnograms were evaluated by computerized automatic analysis using the interval histogram method. 4. An inert placebo was administered on the first 3 nights and on the seventh and eighth nights, and the test article regimen was administered on the fourth, fifth and sixth nights. 5. The test articles and the placebo were administered orally at 22:30 hr, and the recording of polysomnograms was started at 23:00 hr and ended at 8:00 hr the next morning. 6. The subjects were requested to fill out the subjective assessment of sleep before falling asleep and after arising the next morning. 7. BAHS increased REM sleep and decreased stage 2 sleep significantly; however, it failed to affect stage 1, 3 or 4 sleep. 8. Nitrazepam increased significantly the total sleep time and stage 2 sleep but decreased significantly the stage 3 sleep and decreased slightly the stages 1, 4 and REM sleep. 9. The combined treatment with BAHS and nitrazepam did not alter the sleep parameters except for increasing the total sleep time. 10. No obvious changes were observed in the subjective assessments after administration of the drugs. 11. These findings suggest that BAHS results in a unique sleep pattern different from benzodiazepines, and that BAHS may be suitable for treating insomnia in elderly patients and those with drug abuse, manic-depressive illness or schizophrenia.

[Automatic analysis of sleep-wake states during the first year of life]

Automatic analysis on infant sleep tracing was first considered as an aid to the rapid construction of hypnograms. It has long been thought more complicated than in adults, because of the significant changes in electrogenesis during the first year of life, resulting in difficulty to adapt to the criteria of Rechtschaffen and Kales (1968), even if modified, or to those of Anders et al (1971). In fact, these studies have shown that automatic analysis of sleep EEG tracings permit an easier analysis of the changes in electrogenesis according to subject age, sleep stage and time of night. Moreover, automatic analysis of parameters which change with the sleep-wake stages shows the continuity of these changes, and therefore the somewhat arbitrary nature of conventional hypnogram classification. The mathematical treatment of this information facilitates their visualisation, and permits a better analysis of circadian and ultradian variations that could scarcely be formalized by classical hypnogram analysis. In the study of maturation occurring during the first months of life, automatic signal processing will require adaptation that will lead to new forms of reasoning.

Night sleep electroencephalogram power spectral analysis in excessive daytime sleepiness disorders

A group of 53 patients (40 males, 13 females) with mean age of 49 years, ranging from 30 to 70 years, was evaluated in the following excessive daytime sleepiness (EDS) disorders: obstructive sleep apnea syndrome (B4a), periodic movements in sleep (B5a), affective disorder (B2a), functional psychiatric non affective disorder (B2b). We considered all adult patients referred to the Center sequentially with no other distinctions but these three criteria: (a) EDS was the main complaint; (b) right handed; (c) not using psychotropic drugs for two weeks prior to the all-night polysomnography. EEG (C3/A1, C4/A2) samples from 2 to 10 minutes of each stage of the first REM cycle were chosen. The data was recorded simultaneously in magnetic tape and then fed into a computer for power spectral analysis. The percentage of power (PP) in each band calculated in relation to the total EEG power was determined of subsequent sections of 20.4 s for the following frequency bands: delta, theta, alpha and beta. The PP in all EDS patients sample had a tendency to decrease progressively from the slowest to the fastest frequency bands, in every sleep stage. PP distribution in the delta range increased progressively from stage 1 to stage 4; stage REM levels were close to stage 2 levels. In an EDS patients interhemispheric coherence was high in every band and sleep stage. B4a patients sample PP had a tendency to decrease progressively from the slowest to the fastest frequency bands, in every sleep stage; PP distribution in the delta range increased progressively from stage 1 to stage 4; stage REM levels were between stage 1 and stage 2 levels.(ABSTRACT TRUNCATED AT 250 WORDS)