Automatic Detection of Microsleep Episodes With Deep Learning.

Brief fragments of sleep shorter than 15 s are defined as microsleep episodes (MSEs), often subjectively perceived as sleepiness. Their main characteristic is a slowing in frequency in the electroencephalogram (EEG), similar to stage N1 sleep according to standard criteria. The maintenance of wakefulness test (MWT) is often used in a clinical setting to assess vigilance. Scoring of the MWT in most sleep-wake centers is limited to classical definition of sleep (30 s epochs), and MSEs are mostly not considered in the absence of established scoring criteria defining MSEs but also because of the laborious work. We aimed for automatic detection of MSEs with machine learning, i.e., with deep learning based on raw EEG and EOG data as input. We analyzed MWT data of 76 patients. Experts visually scored wakefulness, and according to recently developed scoring criteria MSEs, microsleep episode candidates (MSEc), and episodes of drowsiness (ED). We implemented segmentation algorithms based on convolutional neural networks (CNNs) and a combination of a CNN with a long-short term memory (LSTM) network. A LSTM network is a type of a recurrent neural network which has a memory for past events and takes them into account. Data of 53 patients were used for training of the classifiers, 12 for validation and 11 for testing. Our algorithms showed a good performance close to human experts. The detection was very good for wakefulness and MSEs and poor for MSEc and ED, similar to the low inter-expert reliability for these borderline segments. We performed a visualization of the internal representation of the data by the artificial neuronal network performing best using t-distributed stochastic neighbor embedding (t-SNE). Visualization revealed that MSEs and wakefulness were mostly separable, though not entirely, and MSEc and ED largely intersected with the two main classes. We provide a proof of principle that it is feasible to reliably detect MSEs with deep neuronal networks based on raw EEG and EOG data with a performance close to that of human experts. The code of the algorithms (https://github.com/alexander-malafeev/microsleep-detection) and data (https://zenodo.org/record/3251716) are available.

Automatic detection of microsleep episodes with feature-based machine learning.

STUDY OBJECTIVES: Microsleep episodes (MSEs) are brief episodes of sleep, mostly defined to be shorter than 15 s. In the electroencephalogram (EEG), MSEs are mainly characterized by a slowing in frequency. The identification of early signs of sleepiness and sleep (e.g. MSEs) is of considerable clinical and practical relevance. Under laboratory conditions, the maintenance of wakefulness test (MWT) is often used for assessing vigilance. METHODS: We analyzed MWT recordings of 76 patients referred to the Sleep-Wake-Epilepsy-Center. MSEs were scored by experts defined by the occurrence of theta dominance on ≥1 occipital derivation lasting 1-15 s, whereas the eyes were at least 80% closed. We calculated spectrograms using an autoregressive model of order 16 of 1 s epochs moved in 200 ms steps in order to visualize oscillatory activity and derived seven features per derivation: power in delta, theta, alpha and beta bands, ratio theta/(alpha + beta), quantified eye movements, and median frequency. Three algorithms were used for MSE classification: support vector machine (SVM), random forest (RF), and an artificial neural network (long short-term memory [LSTM] network). Data of 53 patients were used for the training of the classifiers, and 23 for testing. RESULTS: MSEs were identified with a high performance (sensitivity, specificity, precision, accuracy, and Cohen's kappa coefficient). Training revealed that delta power and the ratio theta/(alpha + beta) were most relevant features for the RF classifier and eye movements for the LSTM network. CONCLUSIONS: The automatic detection of MSEs was successful for our EEG-based definition of MSEs, with good performance of all algorithms applied.

Microsleep episodes in the borderland between wakefulness and sleep.

STUDY OBJECTIVES: The wake-sleep transition zone represents a poorly defined borderland, containing, for example, microsleep episodes (MSEs), which are of potential relevance for diagnosis and may have consequences while driving. Yet, the scoring guidelines of the American Academy of Sleep Medicine (AASM) completely neglect it. We aimed to explore the borderland between wakefulness and sleep by developing the Bern continuous and high-resolution wake-sleep (BERN) criteria for visual scoring, focusing on MSEs visible in the electroencephalography (EEG), as opposed to purely behavior- or performance-defined MSEs. METHODS: Maintenance of Wakefulness Test (MWT) trials of 76 randomly selected patients were retrospectively scored according to both the AASM and the newly developed BERN scoring criteria. The visual scoring was compared with spectral analysis of the EEG. The quantitative EEG analysis enabled a reliable objectification of the visually scored MSEs. For less distinct episodes within the borderland, either ambiguous or no quantitative patterns were found. RESULTS: As expected, the latency to the first MSE was significantly shorter in comparison to the sleep latency, defined according to the AASM criteria. In certain cases, a large difference between the two latencies was observed and a substantial number of MSEs occurred between the first MSE and sleep. Series of MSEs were more frequent in patients with shorter sleep latencies, while isolated MSEs were more frequent in patients who did not reach sleep. CONCLUSION: The BERN criteria extend the AASM criteria and represent a valuable tool for in-depth analysis of the wake-sleep transition zone, particularly important in the MWT.

Automatic artefact detection in single-channel sleep EEG recordings.

Quantitative electroencephalogram analysis (e.g. spectral analysis) has become an important tool in sleep research and sleep medicine. However, reliable results are only obtained if artefacts are removed or excluded. Artefact detection is often performed manually during sleep stage scoring, which is time consuming and prevents application to large datasets. We aimed to test the performance of mostly simple algorithms of artefact detection in polysomnographic recordings, derive optimal parameters and test their generalization capacity. We implemented 14 different artefact detection methods, optimized parameters for derivation C3A2 using receiver operator characteristic curves of 32 recordings, and validated them on 21 recordings of healthy participants and 10 recordings of patients (different laboratory) and considered the methods as generalizable. We also compared average power density spectra with artefacts excluded based on algorithms and expert scoring. Analyses were performed retrospectively. We could reliably identify artefact contaminated epochs in sleep electroencephalogram recordings of two laboratories (healthy participants and patients) reaching good sensitivity (specificity 0.9) with most algorithms. The best performance was obtained using fixed thresholds of the electroencephalogram slope, high-frequency power (25-90 Hz or 45-90 Hz) and residuals of adaptive autoregressive models. Artefacts in electroencephalogram data can be reliably excluded by simple algorithms with good performance, and average electroencephalogram power density spectra with artefact exclusion based on algorithms and manual scoring are very similar in the frequency range relevant for most applications in sleep research and sleep medicine, allowing application to large datasets as needed to address questions related to genetics, epidemiology or precision medicine.

Automatic Human Sleep Stage Scoring Using Deep Neural Networks.

The classification of sleep stages is the first and an important step in the quantitative analysis of polysomnographic recordings. Sleep stage scoring relies heavily on visual pattern recognition by a human expert and is time consuming and subjective. Thus, there is a need for automatic classification. In this work we developed machine learning algorithms for sleep classification: random forest (RF) classification based on features and artificial neural networks (ANNs) working both with features and raw data. We tested our methods in healthy subjects and in patients. Most algorithms yielded good results comparable to human interrater agreement. Our study revealed that deep neural networks (DNNs) working with raw data performed better than feature-based methods. We also demonstrated that taking the local temporal structure of sleep into account a priori is important. Our results demonstrate the utility of neural network architectures for the classification of sleep.

The Effect of a Slowly Rocking Bed on Sleep.

Rocking movements appear to affect human sleep. Recent research suggested a facilitated transition from wake to sleep and a boosting of slow oscillations and sleep spindles due to lateral rocking movements during an afternoon nap. This study aimed at investigating the effect of vestibular stimulation on sleep onset, nocturnal sleep and its potential to increase sleep spindles and slow waves, which could influence memory performance. Polysomnography was recorded in 18 males (age: 20-28 years) during three nights: movement until sleep onset (C1), movement for 2 hours (C2), and one baseline (B) without motion. Sleep dependent changes in memory performance were assessed with a word-pair learning task. Although subjects preferred nights with vestibular stimulation, a facilitated sleep onset or a boost in slow oscillations was not observed. N2 sleep and the total number of sleep spindles increased during the 2 h with vestibular stimulation (C2) but not over the entire night. Memory performance increased over night but did not differ between conditions. The lack of an effect might be due to the already high sleep efficiency (96%) and sleep quality of our subjects during baseline. Nocturnal sleep in good sleepers might not benefit from the potential facilitating effects of vestibular stimulation.

Structure and response to flow of the glycocalyx layer.

The glycocalyx is a sugar-rich layer located at the luminal part of the endothelial cells. It is involved in key metabolic processes and its malfunction is related to several diseases. To understand the function of the glycocalyx, a molecular level characterization is necessary. In this article, we present large-scale molecular-dynamics simulations that provide a comprehensive description of the structure and dynamics of the glycocalyx. We introduce the most detailed, to-date, all-atom glycocalyx model, composed of lipid bilayer, proteoglycan dimers, and heparan sulfate chains with realistic sequences. Our results reveal the folding of proteoglycan ectodomain and the extended conformation of heparan sulfate chains. Furthermore, we study the glycocalyx response under shear flow and its role as a flypaper for binding fibroblast growth factors (FGFs), which are involved in diverse functions related to cellular differentiation, including angiogenesis, morphogenesis, and wound healing. The simulations show that the glycocalyx increases the effective concentration of FGFs, leading to FGF oligomerization, and acts as a lever to transfer mechanical stimulus into the cytoplasmic side of endothelial cells.