Computer-assisted sleep staging

Automatic sleep stage classification based on sparse deep belief net and combination of multiple classifiers

This paper presents an automatic sleep stage method combining a sparse deep belief net and combination of multiple classifiers for electroencephalogram, electrooculogram and electromyogram. The sparse deep belief net was applied to extract features from these signals automatically, and the combination of multiple classifiers, utilizing the extracted features, assigned each 30-s epoch to one of the five possible sleep stages. More importantly, we proposed a new voting principle based on classification entropy to enhance the classification performance further by harnessing the complementary information provided by the individual classifier. Differently from existing methods, our method used unsupervised feature learning to extract features automatically from raw sleep data and classification based on the learned features. The results of automatic and manual scorings were compared on an epoch-by-epoch basis. The accuracies for wake, S1, S2, SWS and REM were 98.49%, 80.05%, 91.2%, 98.22% and 95.31%, respectively, and the total accuracy of sleep stage was 91.31%. The results demonstrated that the sparse deep belief net was an efficient feature extraction method for sleep data, and the combination of multiple classifiers based on classification entropy performed well on sleep stages.

A pilot study of serotonin-1A receptor genotypes and rapid eye movement sleep sensitivity to serotonergic/cholinergic imbalance in humans: a pharmacological model of depression

RATIONALE

The serotonergic and cholinergic systems are jointly involved in regulating sleep but this system is theorized to be disturbed in depressed individuals. We previously reported that cholinergic and serotonergic agents induce sleep changes partially consistent with monoamine models of sleep disturbances in depression. One potential cause of disturbed neurotransmission is genetic predisposition. The G(-1019) allele of the serotonin-1A (5-HT1A) receptor promoter region predicts an increased risk for depression compared to the wild-type C(-1019) allele.

OBJECTIVE

The goal of this study was to investigate how serotonin-1A receptor genotypes mediate sleep sensitivity to pharmacological probes modeling the serotonergic/cholinergic imbalance of depression.

METHODS

Seventeen healthy female participants homozygous for either C (n=11) or G (n=6) alleles aged 18-27 years were tested on four nonconsecutive nights. Participants were given galantamine (an anti-acetylcholinesterase), buspirone (a serotonergic agonist), both drugs together, or placebos before sleeping.

RESULTS

As reported previously, buspirone significantly increased rapid eye movement (REM) latency (P<0.001), as well as awakenings, percentage of time spent awake, and percentage of time asleep spent in stage N1 (P<0.019). Galantamine increased awakenings, percentage of time spent awake, percentage of time asleep spent in stage N1, and percentage of time asleep spent in REM, and decreased REM latency and percentage of time asleep spent in stage N3 (P<0.019). Galantamine plus buspirone given together disrupted sleep more than either drug alone, lowering sleep efficiency and percentage of time asleep spent in stage N3 and increasing awakenings, percentage of time spent awake, and percentage of time asleep spent in stage N1 (P<0.019). There was no main effect of genotype nor was there a significant multivariate interaction between genotype and drug condition.

CONCLUSION

These findings are partially consistent with the literature about sleep in depression, notably short REM latency, higher percentage of total sleep time spent in REM, lower percentage of time asleep spent in stage N3, and increased sleep fragmentation. The C/G mutation in the serotonin-1A receptor promoter region does not appear to cause noticeable differences in the sleep patterns of a relatively small sample of healthy young females. Future studies with larger sample sizes are required.

40-Hz ASSR fusion classification system for observing sleep patterns

This paper presents a fusion-based neural network (NN) classification algorithm for 40-Hz auditory steady state response (ASSR) ensemble averaged signals which were recorded from eight human subjects for observing sleep patterns (wakefulness W0 and deep sleep N3 or slow wave sleep SWS). In SWS, sensitivity to pain is the lowest relative to other sleep stages and arousal needs stronger stimuli. 40-Hz ASSR signals were extracted by averaging over 900 sweeps on a 30-s window. Signals generated during N3 deep sleep state show similarities to those produced when general anesthesia is given to patients during clinical surgery. Our experimental results show that the automatic classification system used identifies sleep states with an accuracy rate of 100% when the training and test signals come from the same subjects while its accuracy is reduced to 97.6%, on average, when signals are used from different training and test subjects. Our results may lead to future classification of consciousness and wakefulness of patients with 40-Hz ASSR for observing the depth and effects of general anesthesia (DGA).

The effects of galantamine and buspirone on sleep structure: Implications for understanding sleep abnormalities in major depression

RATIONALE

The serotonergic and cholinergic systems are jointly involved in regulating sleep, but this balance is theorized to be disturbed in depressed individuals.

OBJECTIVE

The goal of this study was to use biological probes in healthy participants, to model the serotonergic/cholinergic imbalance of depression and its associated abnormalities in sleep structure.

METHODS

We tested 20 healthy female participants 18-30 years of age on four non-consecutive nights. Participants were given galantamine (a cholinergic agent), buspirone (a serotonergic agonist), both drugs together, or placebo before sleeping.

RESULTS

Buspirone suppressed tonic rapid eye movement (REM): There was a significant increase in REM latency (p < 0.001). Galantamine increased tonic REM sleep, leading to more time spent in REM (p < 0.001) and shorter REM latency (p < 0.01). Galantamine and buspirone given together were not significantly different from the placebo night by REM sleep measures, but disrupted sleep more than either drug alone.

CONCLUSIONS

These findings are partially consistent with the cholinergic literature about sleep in depression, notably short REM latency, higher percentage of total sleep time spent in REM and increased sleep fragmentation. The prolonged REM latency and reduced percentage of REM with buspirone resembled the effect of selective serotonin reuptake inhibitor antidepressants on REM sleep.

Fusion of Scores in a Detection Context Based on Alpha Integration

We present a new method for fusing scores corresponding to different detectors (two-hypotheses case). It is based on alpha integration, which we have adapted to the detection context. Three optimization methods are presented: least mean square error, maximization of the area under the ROC curve, and minimization of the probability of error. Gradient algorithms are proposed for the three methods. Different experiments with simulated and real data are included. Simulated data consider the two-detector case to illustrate the factors influencing alpha integration and demonstrate the improvements obtained by score fusion with respect to individual detector performance. Two real data cases have been considered. In the first, multimodal biometric data have been processed. This case is representative of scenarios in which the probability of detection is to be maximized for a given probability of false alarm. The second case is the automatic analysis of electroencephalogram and electrocardiogram records with the aim of reproducing the medical expert detections of arousal during sleeping. This case is representative of scenarios in which probability of error is to be minimized. The general superior performance of alpha integration verifies the interest of optimizing the fusing parameters.

Using off-the-shelf lossy compression for wireless home sleep staging

BACKGROUND

Recently, there has been increasing interest in the development of wireless home sleep staging systems that allow the patient to be monitored remotely while remaining in the comfort of their home. However, transmitting large amount of Polysomnography (PSG) data over the Internet is an important issue needed to be considered. In this work, we aim to reduce the amount of PSG data which has to be transmitted or stored, while having as little impact as possible on the information in the signal relevant to classify sleep stages.

NEW METHOD

We examine the effects of off-the-shelf lossy compression on an all-night PSG dataset from 20 healthy subjects, in the context of automated sleep staging. The popular compression method Set Partitioning in Hierarchical Trees (SPIHT) was used, and a range of compression levels was selected in order to compress the signals with various degrees of loss. In addition, a rule-based automatic sleep staging method was used to automatically classify the sleep stages.

RESULTS

Considering the criteria of clinical usefulness, the experimental results show that the system can achieve more than 60% energy saving with a high accuracy (>84%) in classifying sleep stages by using a lossy compression algorithm like SPIHT.

COMPARISON WITH EXISTING METHOD(S)

As far as we know, our study is the first that focuses how much loss can be tolerated in compressing complex multi-channel PSG data for sleep analysis.

CONCLUSIONS

We demonstrate the feasibility of using lossy SPIHT compression for wireless home sleep staging.

A two-step automatic sleep stage classification method with dubious range detection

BACKGROUND

The limitations of the current systems of automatic sleep stage classification (ASSC) are essentially related to the similarities between epochs from different sleep stages and the subjects' variability. Several studies have already identified the situations with the highest likelihood of misclassification in sleep scoring. Here, we took advantage of such information to develop an ASSC system based on knowledge of subjects' variability of some indicators that characterize sleep stages and on the American Academy of Sleep Medicine (AASM) rules.

METHODS

An ASSC system consisting of a two-step classifier is proposed. In the first step, epochs are classified using support vector machines (SVMs) spread into different nodes of a decision tree. In the post-processing step, the epochs suspected of misclassification (dubious classification) are tagged, and a new classification is suggested. Identification and correction are based on the AASM rules, and on misclassifications most commonly found/reported in automatic sleep staging. Six electroencephalographic and two electrooculographic channels were used to classify wake, non-rapid eye movement (NREM) sleep--N1, N2 and N3, and rapid eye movement (REM) sleep.

RESULTS

The proposed system was tested in a dataset of 14 clinical polysomnographic records of subjects suspected of apnea disorders. Wake and REM epochs not falling in the dubious range, are classified with accuracy levels compatible with the requirements for clinical applications. The suggested correction assigned to the epochs that are tagged as dubious enhances the global results of all sleep stages.

CONCLUSIONS

This approach provides reliable sleep staging results for non-dubious epochs.

Learning machines and sleeping brains: Automatic sleep stage classification using decision-tree multi-class support vector machines

BACKGROUND

Sleep staging is a critical step in a range of electrophysiological signal processing pipelines used in clinical routine as well as in sleep research. Although the results currently achievable with automatic sleep staging methods are promising, there is need for improvement, especially given the time-consuming and tedious nature of visual sleep scoring.

NEW METHOD

Here we propose a sleep staging framework that consists of a multi-class support vector machine (SVM) classification based on a decision tree approach. The performance of the method was evaluated using polysomnographic data from 15 subjects (electroencephalogram (EEG), electrooculogram (EOG) and electromyogram (EMG) recordings). The decision tree, or dendrogram, was obtained using a hierarchical clustering technique and a wide range of time and frequency-domain features were extracted. Feature selection was carried out using forward sequential selection and classification was evaluated using k-fold cross-validation.

RESULTS

The dendrogram-based SVM (DSVM) achieved mean specificity, sensitivity and overall accuracy of 0.92, 0.74 and 0.88 respectively, compared to expert visual scoring. Restricting DSVM classification to data where both experts' scoring was consistent (76.73% of the data) led to a mean specificity, sensitivity and overall accuracy of 0.94, 0.82 and 0.92 respectively.

COMPARISON WITH EXISTING METHODS

The DSVM framework outperforms classification with more standard multi-class "one-against-all" SVM and linear-discriminant analysis.

CONCLUSION

The promising results of the proposed methodology suggest that it may be a valuable alternative to existing automatic methods and that it could accelerate visual scoring by providing a robust starting hypnogram that can be further fine-tuned by expert inspection.

A comparative study on classification of sleep stage based on EEG signals using feature selection and classification algorithms

Sleep scoring is one of the most important diagnostic methods in psychiatry and neurology. Sleep staging is a time consuming and difficult task undertaken by sleep experts. This study aims to identify a method which would classify sleep stages automatically and with a high degree of accuracy and, in this manner, will assist sleep experts. This study consists of three stages: feature extraction, feature selection from EEG signals, and classification of these signals. In the feature extraction stage, it is used 20 attribute algorithms in four categories. 41 feature parameters were obtained from these algorithms. Feature selection is important in the elimination of irrelevant and redundant features and in this manner prediction accuracy is improved and computational overhead in classification is reduced. Effective feature selection algorithms such as minimum redundancy maximum relevance (mRMR); fast correlation based feature selection (FCBF); ReliefF; t-test; and Fisher score algorithms are preferred at the feature selection stage in selecting a set of features which best represent EEG signals. The features obtained are used as input parameters for the classification algorithms. At the classification stage, five different classification algorithms (random forest (RF); feed-forward neural network (FFNN); decision tree (DT); support vector machine (SVM); and radial basis function neural network (RBF)) classify the problem. The results, obtained from different classification algorithms, are provided so that a comparison can be made between computation times and accuracy rates. Finally, it is obtained 97.03 % classification accuracy using the proposed method. The results show that the proposed method indicate the ability to design a new intelligent assistance sleep scoring system.

Applications of supervised learning to biological signals: ECG signal quality and systemic vascular resistance

Discovering information encoded in non-invasively recorded biosignals which belies an individual's well-being can help facilitate the development of low-cost unobtrusive medical device technologies, or enable the unsupervised performance of physiological assessments without excessive oversight from trained clinical personnel. Although the unobtrusive or unsupervised nature of such technologies often results in less accurate measures than their invasive or supervised counterparts, this disadvantage is typically outweighed by the ability to monitor larger populations than ever before. The expected consequential benefit will be an improvement in healthcare provision and health outcomes for all. The process of discovering indicators of health in unsupervised or unobtrusive biosignal recordings, or automatically ensuring the validity and quality of such signals, is best realized when following a proven systematic methodology. This paper provides a brief tutorial review of supervised learning, which is a sub-discipline of machine learning, and discusses its application in the development of algorithms to interpret biosignals acquired in unsupervised or semi-supervised environments, with the aim of estimating well-being. Some specific examples in the disparate application areas of telehealth electrocardiogram recording and calculating post-operative systemic vascular resistance are discussed in the context of this systematic approach for information discovery.

Adaptive automatic sleep stage classification under covariate shift

Current automatic sleep stage classification (ASSC) methods that rely on polysomnographic (PSG) signals suffer from inter-subject differences that make them unreliable in facing with new and different subjects. A novel adaptive sleep scoring method based on unsupervised domain adaptation, aiming to be robust to inter-subject variability, is proposed. We assume that the sleep quality variants follow a covariate shift model, where only the sleep features distribution change in the training and test phases. The maximum overlap discrete wavelet transform (MODWT) is applied to extract relevant features from EEG, EOG and EMG signals. A set of significant features are selected by minimum-redundancy maximum-relevance (mRMR) which is a powerful feature selection method. Finally, an instance-weighting method, namely the importance weighted kernel logistic regression (IWKLR) is applied for the purpose of obtaining adaptation in classification. The classification results using leave one out cross-validation (LOOCV), show that the proposed method performs at the state-of-the art in the field of ASSC.

Automatic detection of the wake and stage 1 sleep stages using the EEG sub-epoch approach

Studies by Rechtschaffen and Kales (R&K), rely on 30-sec epochs to score sleep stages. In this paper, we introduce a new approach based on three consecutive and non-consecutive 6-sec sub-epochs for the detection of the wake stage and stage 1 sleep. The Relative Spectral Energy Band (RSEB) is used as a feature extraction from the electroencephalographic (EEG) signal. Spectral estimation is performed using non-parametric and parametric methods. We then compared the performance of the conventional 30-sec epochs with the three consecutive and non-consecutive 6-sec epochs. The outcomes of this study showed that while the accuracy varies between subjects, the non-parametric method proved to be more effective with stage 1 sleep detection and the parametric method was more effective for wake stage detection. The non-consecutive sub-epoch method was more effective and consecutive method was least effective in non-parametric stage 1 detection. Alternatively, the 30-second epoch method was most effective for parametric wake stage detection.

An ensemble system for automatic sleep stage classification using single channel EEG signal

The present work aims at automatic identification of various sleep stages like, sleep stages 1, 2, slow wave sleep (sleep stages 3 and 4), REM sleep and wakefulness from single channel EEG signal. Automatic scoring of sleep stages was performed with the help of pattern recognition technique which involves feature extraction, selection and finally classification. Total 39 numbers of features from time domain, frequency domain and from non-linear analysis were extracted. After extraction of features, SVM based recursive feature elimination (RFE) technique was used to find the optimum number of feature subset which can provide significant classification performance with reduced number of features for the five different sleep stages. Finally for classification, binary SVMs were combined with one-against-all (OAA) strategy. Careful extraction and selection of optimum feature subset helped to reduce the classification error to 8.9% for training dataset, validated by k-fold cross-validation (CV) technique and 10.61% in the case of independent testing dataset. Agreement of the estimated sleep stages with those obtained by expert scoring for all sleep stages of training dataset was 0.877 and for independent testing dataset it was 0.8572. The proposed ensemble SVM-based method could be used as an efficient and cost-effective method for sleep staging with the advantage of reducing stress and burden imposed on subjects.

Automated sleep stage identification system based on time-frequency analysis of a single EEG channel and random forest classifier

In this work, an efficient automated new approach for sleep stage identification based on the new standard of the American academy of sleep medicine (AASM) is presented. The propose approach employs time-frequency analysis and entropy measures for feature extraction from a single electroencephalograph (EEG) channel. Three time-frequency techniques were deployed for the analysis of the EEG signal: Choi-Williams distribution (CWD), continuous wavelet transform (CWT), and Hilbert-Huang Transform (HHT). Polysomnographic recordings from sixteen subjects were used in this study and features were extracted from the time-frequency representation of the EEG signal using Renyi's entropy. The classification of the extracted features was done using random forest classifier. The performance of the new approach was tested by evaluating the accuracy and the kappa coefficient for the three time-frequency distributions: CWD, CWT, and HHT. The CWT time-frequency distribution outperformed the other two distributions and showed excellent performance with an accuracy of 0.83 and a kappa coefficient of 0.76.

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.

A high-performance lossless compression scheme for EEG signals using wavelet transform and neural network predictors

Developments of new classes of efficient compression algorithms, software systems, and hardware for data intensive applications in today's digital health care systems provide timely and meaningful solutions in response to exponentially growing patient information data complexity and associated analysis requirements. Of the different 1D medical signals, electroencephalography (EEG) data is of great importance to the neurologist for detecting brain-related disorders. The volume of digitized EEG data generated and preserved for future reference exceeds the capacity of recent developments in digital storage and communication media and hence there is a need for an efficient compression system. This paper presents a new and efficient high performance lossless EEG compression using wavelet transform and neural network predictors. The coefficients generated from the EEG signal by integer wavelet transform are used to train the neural network predictors. The error residues are further encoded using a combinational entropy encoder, Lempel-Ziv-arithmetic encoder. Also a new context-based error modeling is also investigated to improve the compression efficiency. A compression ratio of 2.99 (with compression efficiency of 67%) is achieved with the proposed scheme with less encoding time thereby providing diagnostic reliability for lossless transmission as well as recovery of EEG signals for telemedicine applications.

A rule-based automatic sleep staging method

In this paper, a rule-based automatic sleep staging method was proposed. Twelve features including temporal and spectrum analyses of the EEG, EOG, and EMG signals were utilized. Normalization was applied to each feature to eliminating individual differences. A hierarchical decision tree with fourteen rules was constructed for sleep stage classification. Finally, a smoothing process considering the temporal contextual information was applied for the continuity. The overall agreement and kappa coefficient of the proposed method applied to the all night polysomnography (PSG) of seventeen healthy subjects compared with the manual scorings by R&K rules can reach 86.68% and 0.79, respectively. This method can integrate with portable PSG system for sleep evaluation at-home in the near future.

Analysis and automatic identification of sleep stages using higher order spectra

Electroencephalogram (EEG) signals are widely used to study the activity of the brain, such as to determine sleep stages. These EEG signals are nonlinear and non-stationary in nature. It is difficult to perform sleep staging by visual interpretation and linear techniques. Thus, we use a nonlinear technique, higher order spectra (HOS), to extract hidden information in the sleep EEG signal. In this study, unique bispectrum and bicoherence plots for various sleep stages were proposed. These can be used as visual aid for various diagnostics application. A number of HOS based features were extracted from these plots during the various sleep stages (Wakefulness, Rapid Eye Movement (REM), Stage 1-4 Non-REM) and they were found to be statistically significant with p-value lower than 0.001 using ANOVA test. These features were fed to a Gaussian mixture model (GMM) classifier for automatic identification. Our results indicate that the proposed system is able to identify sleep stages with an accuracy of 88.7%.

A rule-based automatic sleep staging method

In this paper, a rule-based automatic sleep staging method was proposed. Twelve features, including temporal and spectrum analyses of the EEG, EOG, and EMG signals, were utilized. Normalization was applied to each feature to reduce the effect of individual variability. A hierarchical decision tree, with fourteen rules, was constructed for sleep stage classification. Finally, a smoothing process considering the temporal contextual information was applied for the continuity. The average accuracy and kappa coefficient of the proposed method applied to the all night polysomnography (PSG) of twenty subjects compared with the manual scorings reached 86.5% and 0.78, respectively. This method can assist the clinical staff reduce the time required for sleep scoring in the future.

The use of permutation entropy to characterize sleep electroencephalograms

This work proposes the use of Permutation Entropy (PE), a measure of time-series complexity, to characterize electroencephalogram (EEG) signals recorded during sleep. Such a measure could provide information concerning the different sleep stages and, thus, be utilized as an additional aid to obtain sleep staging information. PE has been estimated for artifact-free 30s segments from more than 80 hours of EEG records obtained from 16 subjects during all-night recordings, from which the mean PE for each sleep stage was obtained. It was found that different sleep stages are characterized by significantly different PE values, which track the physiological changes in the complexity of the EEG signals observed at the different sleep stages. This finding encourages the use of PE as an additional aide to either visual or automated sleep staging.

Quantification of brain macrostates using dynamical nonstationarity of physiological time series

The brain shows complex, nonstationarity temporal dynamics, with abrupt micro- and macrostate transitions during its information processing. Detecting and characterizing these transitions in dynamical states of the brain is a critical issue in the field of neuroscience and psychiatry. In the current study, a novel method is proposed to quantify brain macrostates (e.g., sleep stages or cognitive states) from shifts of dynamical microstates or dynamical nonstationarity. A ``dynamical microstate'' is a temporal unit of the information processing in the brain with fixed dynamical parameters and specific spatial distribution. In this proposed approach, a phase-space-based dynamical dissimilarity map (DDM) is used to detect transitions between dynamically stationary microstates in the time series, and Tsallis time-dependent entropy is applied to quantify dynamical patterns of transitions in the DDM. We demonstrate that the DDM successfully detects transitions between microstates of different temporal dynamics in the simulated physiological time series against high levels of noise. Based on the assumption of nonlinear, deterministic brain dynamics, we also demonstrate that dynamical nonstationarity analysis is useful to quantify brain macrostates (sleep stages I, II, III, IV, and rapid eye movement (REM) sleep) from sleep EEGs with an overall accuracy of 77%. We suggest that dynamical nonstationarity is a useful tool to quantify macroscopic mental states (statistical integration) of the brain using dynamical transitions at the microscopic scale in physiological data.

Artificial neural network and wavelet based automated detection of sleep spindles, REM sleep and wake states

Backpropagation artificial neural network (ANN) has been designed to classify sleep-wake stages. Four hours continuous three channel polygraphic signals such as EEG (electroencephalogram), EOG (electrooculogram) and EMG (electromyogram) from conscious subjects were digitally recorded and stored in computer. EOG and EMG signals were used for manual identification of sleep states before training and testing of ANN. The percentages power of the 2 s epochs of the digitized EEG signals from each of three sleep-wake patterns, sleep spindles (SS), rapid eye movement (REM) sleep and awake (AWA) sates, were calculated and analyzed to select the manually confirmed sleep-wake states for each epoch. Further, second order Daubechies mother wavelet has been used to get the wavelet coefficients for the selected EEG epochs. The wavelet coefficients for the EEG epochs (64 data) were selected as inputs for the training the network and to classify SS, REM sleep and AWA stages. The ANN architecture used (64-14-3) in present study shows overall very good agreement with manual sleep stage scoring with an average of 95.35% for all the 1,140 samples tested from SS, REM and AWA stages. This architecture of ANN was also found effectively differentiating the EEG power spectra from different sleep-wake states (96.84% in SS, 93.68% in REM sleep, 95.52% in AWA state). The high performance observed with the system based on wavelet coefficients along with the ANN, highlights the need of this computational tool into the field of sleep research.

An adaptive error modeling scheme for the lossless compression of EEG signals

Lossless compression of EEG signal is of great importance for the neurological diagnosis as the specialists consider the exact reconstruction of the signal as a primary requirement. This paper discusses a lossless compression scheme for EEG signals that involves a predictor and an adaptive error modeling technique. The prediction residues are arranged based on the error count through an histogram computation. Two optimal regions are identified in the histogram plot through a heuristic search such that the bit requirement for encoding the two regions is minimum. Further improvement in the compression is achieved by removing the statistical redundancy that is present in the residue signal by using a context-based bias cancellation scheme. Three neural network predictors, namely, single-layer perceptron, multilayer perceptron, and Elman network and two linear predictors, namely, autoregressive model and finite impulse response filter are considered. Experiments are conducted using EEG signals recorded under different physiological conditions and the performances of the proposed methods are evaluated in terms of the compression ratio. It is shown that the proposed adaptive error modeling schemes yield better compression results compared to other known compression methods.

Sleep versus wake classification from heart rate variability using computational intelligence: consideration of rejection in classification models

Reliability of classification performance is important for many biomedical applications. A classification model which considers reliability in the development of the model such that unreliable segments are rejected would be useful, particularly, in large biomedical data sets. This approach is demonstrated in the development of a technique to reliably determine sleep and wake using only the electrocardiogram (ECG) of infants. Typically, sleep state scoring is a time consuming task in which sleep states are manually derived from many physiological signals. The method was tested with simultaneous 8-h ECG and polysomnogram (PSG) determined sleep scores from 190 infants enrolled in the collaborative home infant monitoring evaluation (CHIME) study. Learning vector quantization (LVQ) neural network, multilayer perceptron (MLP) neural network, and support vector machines (SVMs) are tested as the classifiers. After systematic rejection of difficult to classify segments, the models can achieve 85%-87% correct classification while rejecting only 30% of the data. This corresponds to a Kappa statistic of 0.65-0.68. With rejection, accuracy improves by about 8% over a model without rejection. Additionally, the impact of the PSG scored indeterminate state epochs is analyzed. The advantages of a reliable sleep/wake classifier based only on ECG include high accuracy, simplicity of use, and low intrusiveness. Reliability of the classification can be built directly in the model, such that unreliable segments are rejected.

Evaluation of automated and semi-automated scoring of polysomnographic recordings from a clinical trial using zolpidem in the treatment of insomnia

OBJECTIVE

To evaluate the performance of 2 automated systems, Morpheus and Somnolyzer24X7, with various levels of human review/editing, in scoring polysomnographic (PSG) recordings from a clinical trial using zolpidem in a model of transient insomnia.

METHODS

164 all-night PSG recordings from 82 subjects collected during 2 nights of sleep, one under placebo and one under zolpidem (10 mg) treatment were used. For each recording, 6 different methods were used to provide sleep stage scores based on Rechtschaffen & Kales criteria: 1) full manual scoring, 2) automated scoring by Morpheus 3) automated scoring by Somnolyzer24X7, 4) automated scoring by Morpheus with full manual review, 5) automated scoring by Morpheus with partial manual review, 6) automated scoring by Somnolyzer24X7 with partial manual review. Ten traditional clinical efficacy measures of sleep initiation, maintenance, and architecture were calculated.

RESULTS

Pair-wise epoch-by-epoch agreements between fully automated and manual scores were in the range of intersite manual scoring agreements reported in the literature (70%-72%). Pair-wise epoch-by-epoch agreements between automated scores manually reviewed were higher (73%-76%). The direction and statistical significance of treatment effect sizes using traditional efficacy endpoints were essentially the same whichever method was used. As the degree of manual review increased, the magnitude of the effect size approached those estimated with fully manual scoring.

CONCLUSION

Automated or semi-automated sleep PSG scoring offers valuable alternatives to costly, time consuming, and intrasite and intersite variable manual scoring, especially in large multicenter clinical trials. Reduction in scoring variability may also reduce the sample size of a clinical trial.

Automatic sleep stage classification using two facial electrodes

Standard sleep stage classification is based on visual analysis of central EEG, EOG and EMG signals. Automatic analysis with a reduced number of sensors has been studied as an easy alternative to the standard. In this study, a single-channel electro-oculography (EOG) algorithm was developed for separation of wakefulness, SREM, light sleep (S1, S2) and slow wave sleep (S3, S4). The algorithm was developed and tested with 296 subjects. Additional validation was performed on 16 subjects using a low weight single-channel Alive Monitor. In the validation study, subjects attached the disposable EOG electrodes themselves at home. In separating the four stages total agreement (and Cohen's Kappa) in the training data set was 74% (0.59), in the testing data set 73% (0.59) and in the validation data set 74% (0.59). Self-applicable electro-oculography with only two facial electrodes was found to provide reasonable sleep stage information.

Sleep-stage scoring in the rat using a support vector machine

Analysis and classification of sleep stages is a fundamental part of basic sleep research. Rat sleep stages are scored based on electrocorticographic (ECoG) signals recorded from electrodes implanted epidurally and electromyographic (EMG) signals from the temporalis or nuchal muscle. An automated sleep scoring system was developed using a support vector machine (SVM) to discriminate among waking, nonrapid eye movement sleep, and paradoxical sleep. Two experts scored retrospective data obtained from six Sprague-Dawley rodents to provide the training sets and subsequent comparison data used to assess the effectiveness of the SVM. Numerous time-domain and frequency-domain features were extracted for each epoch and selectively reduced using statistical analyses. The SVM kernel function was chosen to be a Gaussian radial basis function and kernel parameters were varied to examine the effectiveness of optimization methods. Tests indicated that a common set of features could be chosen resulted in an overall agreement between the automated scores and the expert consensus of greater than 96%.

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.

An automatic analysis method for detecting and eliminating ECG artifacts in EEG

An automated method for detecting and eliminating electrocardiograph (ECG) artifacts from electroencephalography (EEG) without an additional synchronous ECG channel is proposed in this paper. Considering the properties of wavelet filters and the relationship between wavelet basis and characteristics of ECG artifacts, the concepts for selecting a suitable wavelet basis and scales used in the process are developed. The analysis via the selected basis is without suffering time shift for decomposition and detection/elimination procedures after wavelet transformation. The detection rates, above 97.5% for MIT/BIH and NTUH recordings, show a pretty good performance in ECG artifact detection and elimination.

Automatic sleep stage classification using two-channel electro-oculography

An automatic method for the classification of wakefulness and sleep stages SREM, S1, S2 and SWS was developed based on our two previous studies. The method is based on a two-channel electro-oculography (EOG) referenced to the left mastoid (M1). Synchronous electroencephalographic (EEG) activity in S2 and SWS was detected by calculating cross-correlation and peak-to-peak amplitude difference in the 0.5-6 Hz band between the two EOG channels. An automatic slow eye-movement (SEM) estimation was used to indicate wakefulness, SREM and S1. Beta power 18-30 Hz and alpha power 8-12 Hz was also used for wakefulness detection. Synchronous 1.5-6 Hz EEG activity and absence of large eye movements was used for S1 separation from SREM. Simple smoothing rules were also applied. Sleep EEG, EOG and EMG were recorded from 265 subjects. The system was tuned using data from 132 training subjects and then applied to data from 131 validation subjects that were different to the training subjects. Cohen's Kappa between the visual and the developed new automatic scoring in separating 30s wakefulness, SREM, S1, S2 and SWS epochs was substantial 0.62 with epoch by epoch agreement of 72%. With automatic subject specific alpha thresholds for offline applications results improved to 0.63 and 73%. The automatic method can be further developed and applied for ambulatory sleep recordings by using only four disposable, self-adhesive and self-applicable electrodes.