Fine tuned personalized machine learning models to detect insomnia risk based on data from a smart bed platform.

Introduction: Insomnia causes serious adverse health effects and is estimated to affect 10-30% of the worldwide population. This study leverages personalized fine-tuned machine learning algorithms to detect insomnia risk based on questionnaire and longitudinal objective sleep data collected by a smart bed platform. Methods: Users of the Sleep Number smart bed were invited to participate in an IRB approved study which required them to respond to four questionnaires (which included the Insomnia Severity Index; ISI) administered 6 weeks apart from each other in the period from November 2021 to March 2022. For 1,489 participants who completed at least 3 questionnaires, objective data (which includes sleep/wake and cardio-respiratory metrics) collected by the platform were queried for analysis. An incremental, passive-aggressive machine learning model was used to detect insomnia risk which was defined by the ISI exceeding a given threshold. Three ISI thresholds (8, 10, and 15) were considered. The incremental model is advantageous because it allows personalized fine-tuning by adding individual training data to a generic model. Results: The generic model, without personalizing, resulted in an area under the receiving-operating curve (AUC) of about 0.5 for each ISI threshold. The personalized fine-tuning with the data of just five sleep sessions from the individual for whom the model is being personalized resulted in AUCs exceeding 0.8 for all ISI thresholds. Interestingly, no further AUC enhancements resulted by adding personalized data exceeding ten sessions. Discussion: These are encouraging results motivating further investigation into the application of personalized fine tuning machine learning to detect insomnia risk based on longitudinal sleep data and the extension of this paradigm to sleep medicine.

Observational study to understand the effect of timing and regularity on sleep metrics and cardiorespiratory parameters using data from a smart bed.

Sleep regularity and chronotype can affect health, performance, and overall well-being. This observational study examines how sleep regularity and chronotype affect sleep quality and cardiorespiratory metrics. Data was collected from 1 January 2019 through 30 December 2019 from over 330 000 Sleep Number smart bed users across the United States who opted into this at-home study. A pressure signal from the smart bed reflected bed presence, movements, heart rate (HR), and breathing rate (BR). Participants (mean age: 55.69 years [SD: 14.0]; 51.2% female) were categorized by chronotype (16.8% early; 62.2% intermediate, 20.9% late) and regularity of sleep timing. Participants who were regular sleepers (66.1%) experienced higher percent restful sleep and lower mean HR and BR compared to the 4.8% categorized as irregular sleepers. Regular early-chronotype participants displayed better sleep and cardiorespiratory parameters compared to those with regular late-chronotypes. Significant variations were noted in sleep duration (Cohen's d = 1.54 and 0.88, respectively) and restful sleep (Cohen's d = 1.46 and 0.82, respectively) between early and late chronotypes, particularly within regular and irregular sleep patterns. This study highlights how sleep regularity and chronotype influence sleep quality and cardiorespiratory metrics. Irrespective of chronotype, sleep regularity demonstrated a substantial effect. Further research is necessary to confirm these findings.

Electrophysiological model of human temporal contrast sensitivity based on SSVEP.

The present study aims to connect the psychophysical research on the human visual perception of flicker with the neurophysiological research on steady-state visual evoked potentials (SSVEPs) in the context of their application needs and current technological developments. In four experiments, we investigated whether a temporal contrast sensitivity model could be established based on the electrophysiological responses to repetitive visual stimulation and, if so, how this model compares to the psychophysical models of flicker visibility. We used data from 62 observers viewing periodic flicker at a range of frequencies and modulation depths sampled around the perceptual visibility thresholds. The resulting temporal contrast sensitivity curve (TCSC) was similar in shape to its psychophysical counterpart, confirming that the human visual system is most sensitive to repetitive visual stimulation at frequencies between 10 and 20 Hz. The electrophysiological TCSC, however, was below the psychophysical TCSC measured in our experiments for lower frequencies (1-50 Hz), crossed it when the frequency was 50 Hz, and stayed above while decreasing at a slower rate for frequencies in the gamma range (40-60 Hz). This finding provides evidence that SSVEPs could be measured even without the conscious perception of flicker, particularly at frequencies above 50 Hz. The cortical and perceptual mechanisms that apply at higher temporal frequencies, however, do not seem to directly translate to lower frequencies. The presence of harmonics, which show better response for many frequencies, suggests non-linear processing in the visual system. These findings are important for the potential applications of SSVEPs in studying, assisting, or augmenting human cognitive and sensorimotor functions.

Feasibility of Unobtrusively Estimating Blood Pressure Using Load Cells under the Legs of a Bed.

The ability to monitor blood pressure unobtrusively and continuously, even during sleep, may promote the prevention of cardiovascular diseases, enable the early detection of cardiovascular risk, and facilitate the timely administration of treatment. Publicly available data from forty participants containing synchronously recorded signals from four force sensors (load cells located under each leg of a bed) and continuous blood pressure waveforms were leveraged in this research. The focus of this study was on using a deep neural network with load-cell data as input composed of three recurrent layers to reconstruct blood pressure (BP) waveforms. Systolic (SBP) and diastolic (DBP) blood pressure values were estimated from the reconstructed BP waveform. The dataset was partitioned into training, validation, and testing sets, such that the data from a given participant were only used in a single set. The BP waveform reconstruction performance resulted in an R2 of 0.61 and a mean absolute error < 0.1 mmHg. The estimation of the mean SBP and DBP values was characterized by Bland-Altman-derived limits of agreement in intervals of [-11.99 to 15.52 mmHg] and [-7.95 to +3.46 mmHg], respectively. These results may enable the detection of abnormally large or small variations in blood pressure, which indicate cardiovascular health degradation. The apparent contrast between the small reconstruction error and the limit-of-agreement width owes to the fact that reconstruction errors manifest more prominently at the maxima and minima, which are relevant for SBP and DBP estimation. While the focus here was on SBD and DBP estimation, reconstructing the entire BP waveform enables the calculation of additional hemodynamic parameters.

Performance Evaluation of a Smart Bed Technology against Polysomnography.

The Sleep Number smart bed uses embedded ballistocardiography, together with network connectivity, signal processing, and machine learning, to detect heart rate (HR), breathing rate (BR), and sleep vs. wake states. This study evaluated the performance of the smart bed relative to polysomnography (PSG) in estimating epoch-by-epoch HR, BR, sleep vs. wake, mean overnight HR and BR, and summary sleep variables. Forty-five participants (aged 22-64 years; 55% women) slept one night on the smart bed with standard PSG. Smart bed data were compared to PSG by Bland-Altman analysis and Pearson correlation for epoch-by-epoch HR and epoch-by-epoch BR. Agreement in sleep vs. wake classification was quantified using Cohen's kappa, ROC analysis, sensitivity, specificity, accuracy, and precision. Epoch-by-epoch HR and BR were highly correlated with PSG (HR: r = 0.81, |bias| = 0.23 beats/min; BR: r = 0.71, |bias| = 0.08 breaths/min), as were estimations of mean overnight HR and BR (HR: r = 0.94, |bias| = 0.15 beats/min; BR: r = 0.96, |bias| = 0.09 breaths/min). Calculated agreement for sleep vs. wake detection included kappa (prevalence and bias-adjusted) = 0.74 ± 0.11, AUC = 0.86, sensitivity = 0.94 ± 0.05, specificity = 0.48 ± 0.18, accuracy = 0.86 ± 0.11, and precision = 0.90 ± 0.06. For all-night summary variables, agreement was moderate to strong. Overall, the findings suggest that the Sleep Number smart bed may provide reliable metrics to unobtrusively characterize human sleep under real life-conditions.

Interbeat interval-based sleep staging: work in progress toward real-time implementation.

Objective. Cardiac activity changes during sleep enable real-time sleep staging. We developed a deep neural network (DNN) to detect sleep stages using interbeat intervals (IBIs) extracted from electrocardiogram signals.Approach. Data from healthy and apnea subjects were used for training and validation; 2 additional datasets (healthy and sleep disorders subjects) were used for testing. R-peak detection was used to determine IBIs before resampling at 2 Hz; the resulting signal was segmented into 150 s windows (30 s shift). DNN output approximated the probabilities of a window belonging to light, deep, REM, or wake stages. Cohen's Kappa, accuracy, and sensitivity/specificity per stage were determined, and Kappa was optimized using thresholds on probability ratios for each stage versus light sleep.Main results. Mean (SD) Kappa and accuracy for 4 sleep stages were 0.44 (0.09) and 0.65 (0.07), respectively, in healthy subjects. For 3 sleep stages (light+deep, REM, and wake), Kappa and accuracy were 0.52 (0.12) and 0.76 (0.07), respectively. Algorithm performance on data from subjects with REM behavior disorder or periodic limb movement disorder was significantly worse, with Kappa of 0.24 (0.09) and 0.36 (0.12), respectively. Average processing time by an ARM microprocessor for a 300-sample window was 19.2 ms.Significance. IBIs can be obtained from a variety of cardiac signals, including electrocardiogram, photoplethysmography, and ballistocardiography. The DNN algorithm presented is 3 orders of magnitude smaller compared with state-of-the-art algorithms and was developed to perform real-time, IBI-based sleep staging. With high specificity and moderate sensitivity for deep and REM sleep, small footprint, and causal processing, this algorithm may be used across different platforms to perform real-time sleep staging and direct intervention strategies.Novelty & Significance(92/100 words) This article describes the development and testing of a deep neural network-based algorithm to detect sleep stages using interbeat intervals, which can be obtained from a variety of cardiac signals including photoplethysmography, electrocardiogram, and ballistocardiography. Based on the interbeat intervals identified in electrocardiogram signals, the algorithm architecture included a group of convolution layers and a group of long short-term memory layers. With its small footprint, fast processing time, high specificity and good sensitivity for deep and REM sleep, this algorithm may provide a good option for real-time sleep staging to direct interventions.

Heart rate variability increases following automated acoustic slow wave sleep enhancement.

Acoustic stimulation has been shown to enhance slow wave sleep and in turn, cognition, and now cardiac outcomes in young adults. With the emergence of commercial acoustic devices in the home, we sought to examine the impact of an acoustic, slow wave enhancing device on heart rate variability in healthy, middle-aged males (n = 24, 39.92 ± 4.15 years). Under highly controlled conditions, the participants were randomised to receive closed-loop brain state-dependent stimulation in the form of auditory tones (STIM), or no tones (SHAM), in a crossover design, separated by a 1 week washout period. STIM and SHAM were compared on measures of heart rate variability for the whole night and over the first three sleep cycles. We found an increase in slow wave activity following STIM compared with SHAM. There was a significant increase in high frequency power and standard deviation of the normalised RR-intervals (SDNN) during the STIM condition compared with SHAM (p < 0.05), due to changes observed specifically during N3. In conclusion, heart rate variability appears to improve following acoustic slow wave sleep enhancement.

A model characterizing the coupling between slow-wave activity, instantaneous heart rate and heart rate variability during sleep.

The cyclical and progressively decreasing dynamics of electroencephalogram (EEG) based slow-wave activity (SWA) during sleep reflects the homeostatic component of sleep-wake regulation. The dynamic changes of heart rate (HR) and heart rate variability (HRV) indices during sleep also exhibit quasi-cyclic trends that appear to correlate with SWA. This article proposes a model to characterize the relationship between SWA, HR and HRV in the polar-coordinate (r-θ) domain. Polar coordinates are particularly well-suited to model cyclic shapes with simple (linear) equations in the r-θ plane. Group-level analyses and individual-level ones of the correlations between the polar-coordinate transformations of SWA and HR reveal R2 values of 0.99 and 0.95 respectively. Given that, HR and HRV can be estimated in less obtrusive ways compared to EEG. This research offers relevant options to conveniently monitor sleep SWA.Clinical Relevance- Slow wave activity is a marker of sleep restoration that most prominently manifests in the EEG. This research suggests that an electrocardiography (ECG)-based non-linear model can approximate a polar-coordinate version of SWA. Since ECG correlates can be unobtrusively acquired during sleep, these results suggest that practical SWA monitoring can be achieved through cardiac activity measurements.

SSVEP phase synchronies and propagation during repetitive visual stimulation at high frequencies.

Steady-state visual evoked potentials (SSVEPs), the brain response to visual flicker stimulation, have proven beneficial in both research and clinical applications. Despite the practical advantages of stimulation at high frequencies in terms of visual comfort and safety, high frequency-SSVEPs have not received enough attention and little is known about the mechanisms behind their generation and propagation in time and space. In this study, we investigated the origin and propagation of SSVEPs in the gamma frequency band (40-60 Hz) by studying the dynamic properties of EEG in 32 subjects. Using low-resolution brain electromagnetic tomography (sLORETA) we identified the cortical sources involved in SSVEP generation in that frequency range to be in the primary visual cortex, Brodmann areas 17, 18 and 19 with minor contribution from sources in central and frontal sites. We investigated the SSVEP propagation as measured on the scalp in the framework of the existing theories regarding the neurophysiological mechanism through which the SSVEP spreads through the cortex. We found a progressive phase shift from posterior parieto-occipital sites over the cortex with a phase velocity of approx. 8-14 m/s and wavelength of about 21 and 24 cm. The SSVEP spatial properties appear sensitive to input frequency with higher stimulation frequencies showing a faster propagation speed.

Acoustic enhancement of slow wave sleep on consecutive nights improves alertness and attention in chronically short sleepers.

INTRODUCTION: Chronic sleep restriction has been linked to occupational errors and motor vehicle crashes. Enhancing slow wave sleep may alleviate some of the cognitive deficits associated with chronic sleep restriction. However, the extent to which acoustic stimulation of slow wave activity (SWA) may improve alertness and attention is not well established, particularly with respect to consecutive nights of exposure. METHODS: Twenty-five healthy adults (32.9 ± 8.2 years; 16 female) who self-restricted their sleep during workdays participated in a randomized, double-blind, cross-over study. Participants wore an automated acoustic stimulation device for two consecutive nights. Acoustic tones (50 ms long) were delivered on the up-phase of the slow wave first and then at constant 1-s inter-tone-intervals once N3 was identified (STIM), until an arousal or shift to another sleep stage occurred, or at inaudible decibels during equivalent stimulation periods (SHAM). Subjective alertness/fatigue (KSS, Samn-Perelli) was assessed across both days, and objective measures of alertness (MSLT) and attention (PVT) were assessed after two nights of stimulation. RESULTS: After one night of acoustic stimulation, increased slow wave energy was observed in 68% of participants, with an average significant increase of 17.7% (p = 0.01), while Night 2 was associated with a 22.2% increase in SWA (p = 0.08). SWE was highly stable across the two nights of STIM (ICC 0.93, p < 0.001), and around half (56%) of participants were consistently classified as responders (11/25) or non-responders (3/25). Daytime testing showed that participants felt more alert and awake following each night of acoustic stimulation (p < 0.05), with improved objective attention across the day following two nights of acoustic stimulation. DISCUSSION: Consecutive nights of acoustic stimulation enhanced SWA on both nights, and improved next day alertness and attention. Given large individual differences, we highlight the need to examine both the long-term effects of stimulation, and to identify inter-individual differences in acoustic stimulation response. Our findings suggest that the use of an acoustic device to enhance slow wave sleep may alleviate some of the deficits in alertness and attention typically associated with sleep restriction.

Closed-loop Electroencephalogram-based modulated audio to fall and deepen sleep faster.

The transition from wake to sleep is a continuum that is well characterized by the electroencephalogram (EEG) power spectral ratio (ρ) between the beta (15 to 30 Hz) and theta (4 to 8 Hz) bands. From wake to sleep, the value of ρ gradually decreases.We have designed and implemented a single EEG-signal based closed-loop system that leverages ρ to modulate the volume of a pink-noise type of audio such that the volume becomes gradually softer as sleep initiates. A proof-of-concept trial was conducted with this system and it was found that using this concept resulted in a reduction of sleep latency and latency to deep sleep.

Hybrid in-phase and continuous auditory stimulation significantly enhances slow wave activity during sleep.

Recent evidence has shown that enhancing slow-wave activity (SWA) during sleep has positive effects on cognitive, metabolic, and autonomic function. We have developed a consumer, integrated device that automatically detects sleep stages from a single electroencephalogram (EEG) signal and delivers auditory stimulation in a closed-loop manner. The stimulation was delivered in 15-auditory tone blocks separated from each other by at least 15 seconds. The first tone in a block was synchronized to the up-state of a detected slow-wave while subsequent ones were separated from each other by a constant 1-second inter-tone interval. The system was tested in a study involving 22 participants and SWA enhancement (average 45.8%; p=0.0027) was found in 19/22 participants.

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.

Closed-loop system to enhance slow-wave activity.

OBJECTIVE: Recent evidence reports cognitive, metabolic, and sleep restoration benefits resulting from the enhancement of sleep slow-waves using auditory stimulation. Our objective is to make this concept practical for consumer use by developing and validating an electroencephalogram (EEG) closed-loop system to deliver auditory stimulation during sleep to enhance slow-waves. APPROACH: The system automatically detects slow-wave sleep with 74% sensitivity and 97% specificity and optimally delivers stimulation in the form of 50 ms-long tones separated by a constant one-second inter-tone interval at a volume that is dynamically modulated such that louder tones are delivered when sleep is deeper. The system was tested in a study involving 28 participants (18F, 10M; 36.9 ± 7.3 years old; median age: 40 years old) who used the system for ten nights (five nights in a sham condition and five in a stimulation condition). Four nights in each condition were recorded at-home and the fifth one in-lab. MAIN RESULTS: The analysis in two age groups defined by the median age of participants in the study shows significant slow wave activity enhancement (+16.1%, p < 0.01) for the younger group and absence of effect on the older group. However, the older group received only a fraction (57%) of the stimulation compared to the younger group. Changes in sleep architecture and EEG properties due to aging have influenced the amount of stimulation. The analysis of the stimulation timing suggests an entrainment-like phenomenon where slow-waves align to the stimulation periodicity. In addition, enhancement of spindle power in the stimulation condition was found. SIGNIFICANCE: We show evidence of the viability of delivering auditory stimulation during sleep, at home, to enhance slow wave activity. The system ensures the stimulation delivery to be at the right time during sleep without causing disturbance.

Probabilistic characterization of sleep architecture: home based study on healthy volunteers.

The quantification of sleep architecture has high clinical value for diagnostic purposes. While the clinical standard to assess sleep architecture is in-lab based polysomnography, higher ecological validity can be obtained with multiple sleep recordings at home. In this paper, we use a dataset composed of fifty sleep EEG recordings at home (10 per study participant for five participants) to analyze the sleep stage transition dynamics using Markov chain based modeling. The statistical analysis of the duration of continuous sleep stage bouts is also analyzed to identify the speed of transition between sleep stages. This analysis identified two types of NREM states characterized by fast and slow exit rates which from the EEG analysis appear to correspond to shallow and deep sleep respectively.

Neural dynamics during repetitive visual stimulation.

OBJECTIVE: Steady-state visual evoked potentials (SSVEPs), the brain responses to repetitive visual stimulation (RVS), are widely utilized in neuroscience. Their high signal-to-noise ratio and ability to entrain oscillatory brain activity are beneficial for their applications in brain-computer interfaces, investigation of neural processes underlying brain rhythmic activity (steady-state topography) and probing the causal role of brain rhythms in cognition and emotion. This paper aims at analyzing the space and time EEG dynamics in response to RVS at the frequency of stimulation and ongoing rhythms in the delta, theta, alpha, beta, and gamma bands. APPROACH: We used electroencephalography (EEG) to study the oscillatory brain dynamics during RVS at 10 frequencies in the gamma band (40-60 Hz). We collected an extensive EEG data set from 32 participants and analyzed the RVS evoked and induced responses in the time-frequency domain. MAIN RESULTS: Stable SSVEP over parieto-occipital sites was observed at each of the fundamental frequencies and their harmonics and sub-harmonics. Both the strength and the spatial propagation of the SSVEP response seem sensitive to stimulus frequency. The SSVEP was more localized around the parieto-occipital sites for higher frequencies (>54 Hz) and spread to fronto-central locations for lower frequencies. We observed a strong negative correlation between stimulation frequency and relative power change at that frequency, the first harmonic and the sub-harmonic components over occipital sites. Interestingly, over parietal sites for sub-harmonics a positive correlation of relative power change and stimulation frequency was found. A number of distinct patterns in delta (1-4 Hz), theta (4-8 Hz), alpha (8-12 Hz) and beta (15-30 Hz) bands were also observed. The transient response, from 0 to about 300 ms after stimulation onset, was accompanied by increase in delta and theta power over fronto-central and occipital sites, which returned to baseline after approx. 500 ms. During the steady-state response, we observed alpha band desynchronization over occipital sites and after 500 ms also over frontal sites, while neighboring areas synchronized. The power in beta band over occipital sites increased during the stimulation period, possibly caused by increase in power at sub-harmonic frequencies of stimulation. Gamma power was also enhanced by the stimulation. SIGNIFICANCE: These findings have direct implications on the use of RVS and SSVEPs for neural process investigation through steady-state topography, controlled entrainment of brain oscillations and BCIs. A deep understanding of SSVEP propagation in time and space and the link with ongoing brain rhythms is crucial for optimizing the typical SSVEP applications for studying, assisting, or augmenting human cognitive and sensorimotor function.

Automatic characterization of sleep need dissipation dynamics using a single EEG signal.

In the two-process model of sleep regulation, slow-wave activity (SWA, i.e. the EEG power in the 0.5-4 Hz frequency band) is considered a direct indicator of sleep need. SWA builds up during non-rapid eye movement (NREM) sleep, declines before the onset of rapid-eye-movement (REM) sleep, remains low during REM and the level of increase in successive NREM episodes gets progressively lower. Sleep need dissipates with a speed that is proportional to SWA and can be characterized in terms of the initial sleep need, and the decay rate. The goal in this paper is to automatically characterize sleep need from a single EEG signal acquired at a frontal location. To achieve this, a highly specific and reasonably sensitive NREM detection algorithm is proposed that leverages the concept of a single-class Kernel-based classifier. Using automatic NREM detection, we propose a method to estimate the decay rate and the initial sleep need. This method was tested on experimental data from 8 subjects who recorded EEG during three nights at home. We found that on average the estimates of the decay rate and the initial sleep need have higher values when automatic NREM detection was used as compared to manual NREM annotation. However, the average variability of these estimates across multiple nights of the same subject was lower when the automatic NREM detection classifier was used. While this method slightly over estimates the sleep need parameters, the reduced variability across subjects makes it more effective for within subject statistical comparisons of a given sleep intervention.

Enhancement of sleep slow waves: underlying mechanisms and practical consequences.

Even modest sleep restriction, especially the loss of sleep slow wave activity (SWA), is invariably associated with slower electroencephalogram (EEG) activity during wake, the occurrence of local sleep in an otherwise awake brain, and impaired performance due to cognitive and memory deficits. Recent studies not only confirm the beneficial role of sleep in memory consolidation, but also point to a specific role for sleep slow waves. Thus, the implementation of methods to enhance sleep slow waves without unwanted arousals or lightening of sleep could have significant practical implications. Here we first review the evidence that it is possible to enhance sleep slow waves in humans using transcranial direct-current stimulation (tDCS) and transcranial magnetic stimulation. Since these methods are currently impractical and their safety is questionable, especially for chronic long-term exposure, we then discuss novel data suggesting that it is possible to enhance slow waves using sensory stimuli. We consider the physiology of the K-complex (KC), a peripheral evoked slow wave, and show that, among different sensory modalities, acoustic stimulation is the most effective in increasing the magnitude of slow waves, likely through the activation of non-lemniscal ascending pathways to the thalamo-cortical system. In addition, we discuss how intensity and frequency of the acoustic stimuli, as well as exact timing and pattern of stimulation, affect sleep enhancement. Finally, we discuss automated algorithms that read the EEG and, in real-time, adjust the stimulation parameters in a closed-loop manner to obtain an increase in sleep slow waves and avoid undesirable arousals. In conclusion, while discussing the mechanisms that underlie the generation of sleep slow waves, we review the converging evidence showing that acoustic stimulation is safe and represents an ideal tool for slow wave sleep (SWS) enhancement.

Comparison of feature and classifier algorithms for online automatic sleep staging based on a single EEG signal.

Automatic sleep staging on an online basis has recently emerged as a research topic motivated by fundamental sleep research. The aim of this paper is to find optimal signal processing methods and machine learning algorithms to achieve online sleep staging on the basis of a single EEG signal. The classification performance obtained using six different EEG signals and various signal processing feature sets is compared using the kappa statistic which has very recently become popular in sleep staging research. A variable duration of the EEG segment (or epoch) to decide on the sleep stage is also analyzed. Spectral-domain, time-domain, linear, and nonlinear features are compared in terms of performance and two types of machine learning approaches (random forests and support vector machines) are assessed. We have determined that frontal EEG signals, with spectral linear features, epoch durations between 18 and 30 seconds, and a random forest classifier lead to optimal classification performance while ensuring real-time online operation.

Respiratory-cycle related analysis of the EEG-spectrum during sleep: a healthy population study.

Recent research has shown the EEG's spectral changes that occur in synchrony with the respiratory-cycle. During wakefulness, and for healthy subjects it is reported that the EEG power in several frequency bands changes between the expiratory and inspiratory phases. For sleep-disordered breathing (SDB) patients, it is reported that the amplitude of changes in normalized EEG power (referred to as respiratory-cycle related EEG changes RCREC) within a respiratory-cycle decreases after a successful intervention to alleviate the SDB condition. In this paper, we focus on analyzing the changes in the sleep's EEG spectrum related to the respiratory-cycle for a healthy population comprising 39 subjects. For 3 sleep stages (N2, N3, REM), 6 EEG channels, and 7 frequency bands, two types of EEG spectral analyzes were considered: 1) the ratio between the EEG power during expiration and that during inspiration, and 2) the RCREC. For the first type of analysis and at the population level, no statistically significant difference was found between the EEG power during expiration and that during inspiration. For the second type of analysis, the RCREC for all conditions is at a level that is statistically significantly larger than 0.1. The latter being the value at which the RCREC decreased after successful SDB intervention.