Sleep Stages Research Articles

The impact of sleep disturbances on treatment efficacy and prognosis in patients with depressive and anxiety disorders.

Little was known about the relationship between sleep disturbances and depressive and anxiety disorders, as well as the efficacy of treatment regimens. During 2021 to 2023, a total of 417 participants were screened by Hamilton Depression Rating Scale (HAMD-17) and Hamilton Anxiety Rating Scale (HAMA-14) for psychological status, and Pittsburgh sleep quality index (PSQI) assessment. 409 participants were finally enrolled, of which 188 (45.97%) were suffered from sleep disorders. All participants were received polysomnography (PSG) followed by six-week pharmacological treatment of escitalopram and zopiclone, and finally assessed by HAMD-17 and HAMA-14 for treatment efficacy. PSG monitoring indicated that participants with depression experienced prolonged rapid eye movement sleep latency (REMSL) and increased wakefulness after sleep onset (WASO) (P=0.030 and P=0.002, respectively). Those with anxiety disorders demonstrated a significantly higher percentage of non-rapid eye movement sleep (NREM%) and reduced WASO (P=0.013 and P=0.001, respectively). After six-weeks pharmacological treatment, participants with or without sleep disorders exhibited with similar efficacy outcomes of depression and anxiety disorders (P>0.05). However, every point of PSQI increment at baseline would decrease 0.78 and 0.85 times of probability of effective pharmacological treatment of depression and anxiety disorders. Moreover, participants with both effective outcomes of depression and anxiety disorders were found significant shorter sleep onset latency (SOL) (P<0.001). The insights gained underscore the necessity of considering sleep disturbances in enhancing the overall effectiveness of pharmacological treatments for depression and anxiety disorders.

Recent Evidence of Non-Rapid Eye Movement Sleep Oscillation Abnormalities in Psychiatric Disorders.

We review recent studies published from 2019 to 2024 examining slow waves and sleep spindles abnormalities across neurodevelopmental, mood, trauma-related, and psychotic disorders using polysomnography and Electroencephalogram (EEG). Individuals with attention-deficit/hyperactivity disorder (ADHD) showed higher slow-spindle activity, while findings on slow-wave activity were mixed. Individuals with autism spectrum disorder (ASD) showed inconsistent results with some evidence of lower spindle chirp and slow-wave amplitude. Individuals with depression displayed lower slow-wave and spindle parameters mostly in medicated patients. Individuals with post-traumatic stress disorder (PTSD) showed higher spindle frequency and activity, which were associated with their clinical symptoms. Psychotic disorders demonstrated the most consistent alterations, with lower spindle density, amplitude, and duration across illness stages that correlated with patients' symptom severity and cognitive deficits, whereas lower slow-wave measures were present in the early phases of the disorders. Sleep spindle and slow-wave abnormalities are present across psychiatric populations, with the most consistent alterations observed in psychotic disorders. Larger studies with standardized methodologies and longitudinal assessments are needed to establish the potential of these oscillations as neurophysiological biomarkers and/or treatment targets.

Heart Rate Variability During Rapid Eye Movement Sleep is Associated with Reduced Negative Memory Bias.

Emotional memories change over time, but the mechanisms supporting this change are not well understood. Sleep has been identified as one mechanism that supports memory consolidation, with sleep selectively benefitting negative emotional consolidation at the expense of neutral memories, with specific oscillatory events linked to this process. In contrast, the consolidation of neutral and positive memories, compared to negative memories, has been associated with increased vagally-mediated vagal heart rate variability (HRV) during wakefulness. However, how HRV during sleep contributes to emotional memory consolidation remains unexplored. We investigated how sleep oscillations and vagal contributions during sleep contribute to the consolidation of neutral and negative memories. Using a double-blind, placebo-controlled, within-subject, cross-over design, we examined the impact of pharmacological vagal suppression using zolpidem on overnight emotional memory consolidation. Thirty-two participants encoded neutral and negative pictures in the morning, followed by picture recognition tests before and after a night of sleep. Zolpidem or placebo was administered in the evening before overnight sleep, and participants were monitored with electroencephalography and electrocardiography. In the placebo condition, greater overnight improvement for neutral pictures was associated with higher vagal HRV in both Non-Rapid Eye Movement Slow Wave Sleep (NREM SWS) and REM. Additionally, the emotional memory tradeoff (i.e., difference between consolidation of neutral versus negative memories) was associated with higher vagal HRV during REM, but in this case, neutral memories were remembered better than negative memories, indicating a potential role for REM vagal HRV in promoting a positive memory bias overnight. Zolpidem, on the other hand, reduced vagal HRV during SWS, increased NREM sigma power, and eliminated the positive memory bias. Lastly, we used a stepwise linear mixed effects regression to determine how NREM sigma power and vagal HRV during REM independently explained the variance in the emotional memory tradeoff effect. We found that the addition of vagal HRV in combination with sleep significantly improved the model's fit. Overall, our results suggest that sleep brain oscillations and vagal signals synergistically interact in the overnight consolidation of emotional memories, with REM vagal HRV critically contributing to the positive memory bias.

Study on cerebral oxygen saturation in children with sleep-disordered breathing.

To explore the association between the severity of sleep-disordered breathing, different types of respiratory events, peripheral oxygen saturation (SpO2), age and sleep stage on cerebral oxygen saturation (rSO2) in children. We enrolled children aged 4-14 years who were treated for snoring or mouth breathing at the Sleep Center of Beijing Children's Hospital, from February 2022 to July 2022. All children completed polysomnography, and SpO2, rSO2, and heart rate (HR) were recorded synchronously. A total of 70 children were included, including 16 (22.9%) with primary snoring, 38 (54.3%) with mild obstructive sleep apnea (OSA), and 16 (22.9%) with moderate-to-severe OSA. There were no significant differences in the mean rSO2 or minimum rSO2 among the primary snoring, mild OSA, and moderate-to-severe OSA groups (all p > 0.05). A total of 1119 respiratory events were included in the analysis. Regardless of the type of respiratory event, rSO2 and HR changes occur prior to fluctuations in SpO2. A mixed-effects model showed that ΔrSO2 was positively correlated with ΔSpO2, duration of respiratory event, mixed and obstructive apnea, central apnea, while negatively correlated with age and rapid eye movement (REM) sleep stage (all p < 0.05). Larger rSO2 fluctuations were impacted by a greater ΔSpO2, longer duration of respiratory events, younger age, apnea-related respiratory events and non-REM sleep stage. Thus, sleep disordered breathing in younger children warrants more attention. More research is needed to determine whether REM sleep has special protective effects on rSO2.

Automatic Classification of Sleep Stages from EEG Signals Using Riemannian Metrics and Transformer Networks

Automatic Classification of Sleep Stages from EEG Signals Using Riemannian Metrics and Transformer Networks

Probabilistic Sleep Staging in MSLTs across Hypersomnia Disorders.

This study aimed to identify novel markers of narcolepsy type 1 (NT1) using between-nap opportunity periods ('Lights On') and in-nap opportunity periods ('Lights Off') features of Multiple Sleep Latency Test (MSLT) recordings. We hypothesized that NT1 could be identified both from sleep-wake instability and patterns of sleepiness during wakefulness. Further, we explored if MSLTs from NT1 and narcolepsy type 2 (NT2) patients could be distinguished despite having the same diagnostic thresholds. We analyzed 'Lights On' and 'Lights Off' periods of the MSLT, extracting 163 features describing sleepiness, microsleep, and sleep stage mixing using data from 177 patients with NT1, NT2, Idiopathic Hypersomnia (IH), and Subjective Hypersomnia (sH) from three sleep centers. These features were based on automated probabilistic sleep staging, also denoted as hypnodensities, using U-Sleep. Hypersomnias were differentiated using either or both features from 'Lights On' and 'Lights Off'. Patients with NT1 could be distinguished from NT2, IH, and sH using features solely from 'Lights On' periods with a sensitivity of 0.76 and specificity of 0.71. When using features from all periods of the MSLT, NT1 was distinguished from NT2 alone with a sensitivity of 0.77 and a specificity of 0.84. The findings of this study demonstrate microsleeps and sleep stage mixing as potential markers of the sleep attacks and unstable sleep-wake states common in NT1. Further, NT1 and NT2 could be frequently distinguished using 'Lights Off' features.

Caffeine Intake Alters Recovery Sleep after Sleep Deprivation.

Caffeine is a well-known psychostimulant reputed to alleviate the deleterious effects of sleep deprivation. Nevertheless, caffeine can alter sleep duration and quality, particularly during recovery sleep. We evaluated the effects of acute caffeine intake on the duration and quality of recovery sleep following total sleep deprivation (TSD), taking into account daily caffeine consumption. Forty-one participants performed a double-blind, crossover TSD protocol (38 h of continuous wakefulness) with acute caffeine or placebo. Caffeine (2.5 mg/kg) or placebo was administered twice during continuous wakefulness (last treatment 6.5 h before bedtime for the recovery night). Polysomnographic measurements were recorded using a connected headband. TSD was associated with a rebound in total sleep time (TST) on the recovery night (+110.2 ± 23.2 min, p < 0.001). Caffeine intake decreased this recovery TST (-30.2 ± 8.2 min p = 0.02) and the N3 sleep stage duration (-35.6 ± 23.2 min, p < 0.01). Caffeine intake altered recovery sleep continuity (increased number of long awakenings), stability (higher stage transition frequency), and organization (less time spent in complete sleep cycle) and decreased the delta power spectral density during NREM sleep. On the recovery night, habitual daily caffeine consumption was negatively correlated with TST in caffeine and placebo conditions and positively correlated with wake after sleep onset (WASO) duration and with the frequency of long (>2 min) awakenings in the caffeine condition only. Acute caffeine intake during TSD affects nighttime recovery sleep, with an interaction with daily consumption. These results may influence advice on caffeine intake for night-shift workers. (NCT03859882).

Optimal sizing of battery energy storage system (BESS) for multiple applications using regression analysis and deep sleep optimizer algorithm

The multifunctional applications of battery energy storage system in a power system network will reduce the significant slack time of use as evident in many single-based applications. In order to deploy BESS for multiple applications, it is of utmost importance that the optimal size for the desired multiple functions, firstly be determined. This work proposes a novel methodology for the optimal sizing of battery energy storage system for frequency support, power loss minimization and voltage deviation mitigations. The suggested sizing methodology takes into account the level of penetration of the renewable energy sources in the power network. Regression analysis is used for mathematical formulations while Deep Sleep Heuristic algorithms in MATLAB environment is used for the optimization process for BESS optimal size. The robustness of the proposed method was tested by using IEEE modified 39-bus system. Simulation results show that with the BESS optimal size integrated into the network, voltage deviations were mitigated by about 20 % and power losses were reduced from 65.3 MW to 59.68 MW. Also, the system frequency nadir during the outage of the largest single generating, was sustained at 59.60 Hz whereas, without BESS it was 59.15 Hz. This is critical because such a frequency decline may activate the underfrequency load shedding relays.

Camphor alters occipital electrocorticographic patterns during sleep deprivation in Wistar rats

IntroductionSleep disorders are common in the general population, necessitating the search for new strategies to address this public health challenge. The study aims to describe the electrocorticographic and behavioral changes in sleep deprived Wistar rats exposed to varying doses of camphor, to assess its effects on sleep and its potential as a sleep-inducing drug.Materials and MethodsFor the electrocorticographic evaluation, seventy-two rats were randomly assigned to distinct groups: a control group, a sleep-deprived group, three sleep-deprived groups receiving 10, 20, and 30 mg/kg i.p. of camphor respectively, and three groups that received these doses without sleep deprivation. For the behavioral analysis, twenty-seven rats were divided into three groups, each receiving the same doses as the previous test.Results and DiscussionOur results showed that there was a decrease in the frequency of brain oscillatory patterns when camphor was administered at 10 mg/kg i.p. whereas there was a dose-dependent increase in the spectral power and distribution following the administration of 20 and 30 mg/kg i.p., with the emergence of Delta, Theta, Alpha, and Beta waves. As for the behavioral analysis, it was demonstrated that testicular relaxation, decreased motility, and light sleep induction also occurred in a dose-dependent manner. Thus, we conclude that camphor administration intensifies occipital electrocorticographic patterns in sleep-deprived rats, and its electrocorticographic and behavioral analysis could indicate a potential as a supporting agent in the insomnia treatment.

Impact of background input on memory consolidation

Memory consolidation involves repeated replay of new information by the hippocampus, which transfers memories to the neocortex for long-term storage. This occurs mainly during slow wave sleep, a phase characterized in the cortex by low cholinergic tone and low afferent input. High cholinergic tone has been shown to hamper memory consolidation, probably mediated by reduced network excitability (the ease of activity propagation in a network). We used cortical neuronal networks on multi electrode arrays to investigate whether low background input contributes to memory consolidation. Networks received focal electrical stimuli to memorize, with or without background afferent input (global optogenetic stimulation). Background stimulation hampered memory formation and consolidation, confirming the importance of low background input. Moreover, it lowered network excitability, similar to high cholinergic tone. These findings suggest that high network excitability is a critical feature of slow wave sleep that facilitates memory consolidation.

Selective Modulation of Fear Memory in Non-Rapid Eye Movement Sleep.

Sleep stabilizes memories for their consolidation, but how to modify specific fear memory during sleep remains unclear. Here, it is reported that using targeted memory reactivation (TMR) to reactivate prior fear learning experience in non-slow wave sleep (NS) inhibits fear memory consolidation, while TMR during slow wave sleep (SWS) enhances fear memory in mice. Replaying conditioned stimulus (CS) during sleep affects sleep spindle occurrence, leading to the reduction or enhancement of slow oscillation-spindle (SO-spindle) coupling in NS and SWS, respectively. Optogenetic inhibition of pyramidal neurons in the frontal association cortex (FrA) during TMR abolishes the behavioral effects of NS-TMR and SWS-TMR by modulating SO-spindle coupling. Notably, calcium imaging of the L2/3 pyramidal neurons in the FrA shows that CS during SWS selectively enhances the activity of neurons previously activated during fear conditioning (FC+ neurons), which significantly correlates with CS-elicited spindle power spectrum density. Intriguingly, these TMR-induced calcium activity changes of FC+ neurons further correlate with mice freezing behavior, suggesting their contributions to the consolidation of fear memories. The findings indicate that TMR can selectively weaken or strengthen fear memory, in correlation with modulating SO-spindle coupling and the reactivation of FC+ neurons during substages of non-rapid eye movement (NREM) sleep.

Rapid assessment of cardiac autonomic modulation and adaptive stress responses: Automatic calculation of time-varying parasympathetic, sympathetic, and Baevsky stress indexes

Cardiac autonomic modulation (CAM), which is regulated by the balance between the sympathetic and parasympathetic nervous systems, is involved in various physiological and pathological conditions. Heart rate variability (HRV) analysis has been used to explore the complex relationship between the brain and heart, as described by Porges&amp;rsquo; polyvagal theory and Thayer&amp;rsquo;s neurovisceral integration model. Recently, an automated calculation of new parasympathetic, sympathetic, and Baevsky stress indexes based on HRV parameters has been introduced for faster and more comprehensive CAM assessment, though their normal ranges remain undefined. This study aimed to determine the average values of these indexes in a healthy population of different ages during rest, daily activities, non-rapid eye movement sleep, graded physical effort, and acute psychophysiological stress. At rest, the parasympathetic and sympathetic indexes were consistently within the proposed normal range and inversely related. However, Baevsky stress index values from Kubios were higher than expected, conflicting with the assumption that they are simply the square root of those calculated using the original formula. Despite this, time-varying assessment of all indexes can provide valuable insights into CAM adaptation during physical effort and acute psychophysiological stress in real-world critical situations. Notably, our novel finding shows that the inverse correlation between parasympathetic and sympathetic/stress indexes under stress is better explained by non-linear functions, offering a potential new measure of brain&amp;ndash;heart interaction during real-life critical events.

Impact of sleep disturbances on outcomes in intensive care units

BackgroundSleep deprivation is common in intensive care units (ICUs) and may alter respiratory performance. Few studies have assessed the role of sleep disturbances on outcomes in critically ill patients.ObjectivesWe hypothesized that sleep disturbances may be associated with poor outcomes in ICUs.MethodsPost-hoc analysis pooling three observational studies assessing sleep by complete polysomnography in 131 conscious and non-sedated patients included at different times of their ICU stay. Sleep was assessed early in a group of patients admitted for acute respiratory failure while breathing spontaneously (n = 34), or under mechanical ventilation in patients with weaning difficulties (n = 45), or immediately after extubation (n = 52). Patients admitted for acute respiratory failure who required intubation, those under mechanical ventilation who had prolonged weaning, and those who required reintubation after extubation were considered as having poor clinical outcomes. Durations of deep sleep, rapid eye movement (REM) sleep, and atypical sleep were compared according to the timing of polysomnography and the clinical outcomes.ResultsWhereas deep sleep remained preserved in patients admitted for acute respiratory failure, it was markedly reduced under mechanical ventilation and after extubation (p < 0.01). Atypical sleep was significantly more frequent in patients under mechanical ventilation than in those breathing spontaneously (p < 0.01). REM sleep was uncommon at any time of their ICU stay. Patients with complete disappearance of REM sleep (50% of patients) were more likely to have poor clinical outcomes than those with persistent REM sleep (24% vs. 9%, p = 0.03).ConclusionComplete disappearance of REM sleep was significantly associated with poor clinical outcomes in critically ill patients.

A double-layered fully automated insomnia identification model employing synthetic data generation using MCSA and CTGAN with single-channel EEG signals.

Insomnia was diagnosed by analyzing sleep stages obtained during polysomnography (PSG) recording. The state-of-the-art insomnia detection models that used physiological signals in PSG were successful in classification. However, the sleep stages of unbalanced data in small-time intervals were fed for classification in previous studies. This can be avoided by analyzing the insomnia detection structure in different frequency bands with artificially generated data from the existing one at the preprocessing and post-processing stages. Hence, the paper proposes a double-layered augmentation model using Modified Conventional Signal Augmentation (MCSA) and a Conditional Tabular Generative Adversarial Network (CTGAN) to generate synthetic signals from raw EEG and synthetic data from extracted features, respectively, in creating training data. The presented work is independent of sleep stage scoring and provides double-layered data protection with the utility of augmentation methods. It is ideally suited for real-time detection using a single-channel EEG provides better mobility and comfort while recording. The work analyzes each augmentation layer's performance individually, and better accuracy was observed when merging both. It also evaluates the augmentation performance in various frequency bands, which are decomposed using discrete wavelet transform, and observed that the alpha band contributes more to detection. The classification is performed using Decision Tree (DT), Ensembled Bagged Decision Tree (EBDT), Gradient Boosting (GB), Random Forest (RF), and Stacking classifier (SC), attaining the highest classification accuracy of 94% using RF with a greater Area Under Curve (AUC) value of 0.97 compared to the existing works and is best suited for small datasets.

Cardiovascular and vasomotor pulsations in the brain and periphery during awake and NREM sleep in a multimodal fMRI study.

The cerebrospinal fluid dynamics in the human brain are driven by physiological pulsations, including cardiovascular pulses and very low-frequency (< 0.1 Hz) vasomotor waves. Ultrafast functional magnetic resonance imaging (fMRI) facilitates the simultaneous measurement of these signals from venous and arterial compartments independently with both classical venous blood oxygenation level dependent (BOLD) and faster arterial spin-phase contrast. In this study, we compared the interaction of these two pulsations in awake and sleep using fMRI and peripheral fingertip photoplethysmography in both arterial and venous signals in 10 healthy subjects (5 female). Sleep increased the power of brain cardiovascular pulsations, decreased peripheral pulsation, and desynchronized them. However, vasomotor waves increase power and synchronicity in both brain and peripheral signals during sleep. Peculiarly, lag between brain and peripheral vasomotor signals reversed in sleep within the default mode network. Finally, sleep synchronized cerebral arterial vasomotor waves with venous BOLD waves within distinct parasagittal brain tissue. These changes in power and pulsation synchrony may reflect systemic sleep-related changes in vascular control between the periphery and brain vasculature, while the increased synchrony of arterial and venous compartments may reflect increased convection of regional neurofluids in parasagittal areas in sleep.

Relationship Between High Frequency Component of Heart Rate Variability and Delta EEG Power During Sleep in Women With Irritable Bowel Syndrome Compared to Healthy Women.

To explore the relationship between the high frequency (HF) heart rate variability (HRV) and electroencephalogram (EEG) delta band power in women with irritable bowel syndrome (IBS) versus healthy control women. Twenty women with IBS and twenty healthy controls were studied over three consecutive nights using polysomnography in a sleep laboratory. To avoid the first night effect, only second-night data were analyzed. Power spectral analysis was applied to HRV and EEG recordings. The linear system coherence/phase analysis assessed the relationship between normalized HF power of HRV and normalized delta band power of EEG during the first four NREM-REM sleep cycles. Women with IBS exhibited a significantly higher percentage of NREM sleep, higher normalized HF, lower normalized low frequency (LF) and decreased LF/HF ratio of HRV in the first four NREM-REM sleep cycles compared to controls. Additionally, their normalized delta band power was significantly lower in these sleep cycles and over the whole night. The phase shift between HF and delta band power was significantly longer in the IBS group. While the coherence between normalized HF and normalized delta band power was lower in the IBS group, the difference was not statistically significant. The coherence/phase analysis showed a dysregulated interaction between autonomic and central nervous systems in women with IBS, manifested by increased lag time between cardiac and EEG delta band power compared to healthy controls. Whether this dysregulation contributes to the pathophysiology of IBS remains to be determined.

Pichia pastoris secreted peptides crossing the blood-brain barrier and DSIP fusion peptide efficacy in PCPA-induced insomnia mouse models.

Pichia pastoris-secreted delta sleep inducing peptide and crossing the blood-brain barrier peptides (DSIP-CBBBP) fusion peptides holds significant promise for its potential sleep-enhancing and neurotransmitter balancing effects. This study investigates these properties using a p-chlorophenylalanine (PCPA) -induced insomnia model in mice, an approach akin to traditional methods evaluating sleep-promoting activities in fusion peptides. The research aims to elucidate the sleep-promoting mechanism of DSIP-CBBBP, exploring its impact on neurotransmitter levels and sleep regulation, and to analyze its composition and structure. Using a PCPA-induced insomnia mouse model, the study evaluates the sleep-promoting effects of DSIP-CBBBP. The peptide's influence on neurotransmitters such as 5-HT, glutamate, dopamine, and melatonin is assessed. The functions of DSIP-CBBBP are characterized using biochemical and animal insomnia-induced behavior tests and compared without CBBBP. DSIP-CBBBP demonstrates a capacity to modulate neurotransmitter levels, indicated by changes in 5-HT, glutamate, DA, and melatonin. DSIP-CBBBP shows a better restorative effect than DSIP on neurotransmitter imbalance and the potential to enhance sleep. The study underscores DSIP-CBBBP potential in correcting neurotransmitter dysregulation and promoting sleep, hinting at its utility in sleep-related therapies.

Sleep deprivation-induced memory impairment: exploring potential interventions.

Sleep's crucial role in memory consolidation is well-established, with neuroimaging and sleep stage analysis revealing the intricate processes involved. Sleep deprivation significantly impairs memory performance and the ability to form new memories, highlighting the need for effective countermeasures. This article concludes that while sleep deprivation significantly impairs memory, the emerging insights into the gut-brain axis offer a promising frontier for developing novel interventions that can mitigate these effects. The review discusses various interventions, ranging from pharmaceutical compounds like donepezil, memantine, and tolcapone, to innovative physical therapy techniques such as transcranial magnetic stimulation (TMS), deep brain stimulation (DBS), and transcranial direct current stimulation (tDCS). Additionally, the emerging role of the gut-brain axis in sleep deprivation-induced memory impairment is examined, shedding light on the complex interplay between gut microbiota and cognitive functions. This comprehensive review explores the multifaceted relationship between sleep deprivation and memory impairment, delving into the neurobiological mechanisms underlying these processes and examining potential interventions.

Structure of Spectral Composition and Synchronization in Human Sleep on the Whole Scalp: A Pilot Study.

We used numerical methods to define the normative structure of the different stages of sleep and wake (W) in a pilot study of 19 participants without pathology (18-64 years old) using a double-banana bipolar montage. Artefact-free 120-240 s epoch lengths were visually identified and divided into 1 s windows with a 10% overlap. Differential channels were grouped into frontal, parieto-occipital, and temporal lobes. For every channel, the power spectrum (PS) was calculated via fast Fourier transform and used to compute the areas for the delta (0-4 Hz), theta (4-8 Hz), alpha (8-13 Hz), and beta (13-30 Hz) bands, which were log-transformed. Furthermore, Pearson's correlation coefficient and coherence by bands were computed. Differences in logPS and synchronization from the whole scalp were observed between the sexes for specific stages. However, these differences vanished when specific lobes were considered. Considering the location and stages, the logPS and synchronization vary highly and specifically in a complex manner. Furthermore, the average spectra for every channel and stage were very well defined, with phase-specific features (e.g., the sigma band during N2 and N3, or the occipital alpha component during wakefulness), although the slow alpha component (8.0-8.5 Hz) persisted during NREM and REM sleep. The average spectra were symmetric between hemispheres. The properties of K-complexes and the sigma band (mainly due to sleep spindles-SSs) were deeply analyzed during the NREM N2 stage. The properties of the sigma band are directly related to the density of SSs. The average frequency of SSs in the frontal lobe was lower than that in the occipital lobe. In approximately 30% of the participants, SSs showed bimodal components in the anterior regions. qEEG can be easily and reliably used to study sleep in healthy participants and patients.

Unravelling sleep patterns: Supervised contrastive learning with self-attention for sleep stage classification

Sleep data scoring is a crucial and primary step for diagnosing sleep disorders to know the sleep stages from the PSG signals. This study uses supervised contrastive learning with a self-attention mechanism to classify sleep stages. We propose a deep learning framework for automatic sleep stage classification, which involves two training phases: (1) the feature representation learning phase, in which the feature representation network (encoder) learns to extract features from the electroencephalogram (EEG) signals, and (2) the classification network training phase, where a pre-trained encoder (trained during phase I) along with the classifier head is fine-tuned for the classification task. The PSG data shows a non-uniform distribution of sleep stages, with wake (W) (around 30% of total samples) and N2 stages (around 58% and 37% of total samples in Physionet EDF-Sleep 2013 and 2018 datasets, respectively) being more prevalent, leading to an imbalanced dataset. The imbalanced data issue is addressed using a weighted softmax cross-entropy loss function that assigns higher weights to minority sleep stages. Additionally, an oversampling technique (the synthetic minority oversampling technique (SMOTE) (Chawla et al., 2002)[1] ) is applied to generate synthetic samples for minority classes. The proposed model is evaluated on the Physionet EDF-Sleep 2013 and 2018 datasets using Fpz-Cz and Pz-Oz EEG channels. It achieved an overall accuracy of 94.1%, a macro F1 score of 92.64, and a Cohen’s Kappa coefficient of 0.92. Ablation studies demonstrated the importance of triplet loss-based pre-training and oversampling for enhancing performance. The proposed model requires minimal pre-processing, eliminating the need for extensive signal processing expertise, and thus is well-suited for clinicians diagnosing sleep disorders.