U-Sleep: resilient high-frequency sleep staging.

Sleep disorders affect a large portion of the global population and are strong predictors of morbidity and all-cause mortality. Sleep staging segments a period of sleep into a sequence of phases providing the basis for most clinical decisions in sleep medicine. Manual sleep staging is difficult and time-consuming as experts must evaluate hours of polysomnography (PSG) recordings with electroencephalography (EEG) and electrooculography (EOG) data for each patient. Here, we present U-Sleep, a publicly available, ready-to-use deep-learning-based system for automated sleep staging ( sleep.ai.ku.dk ). U-Sleep is a fully convolutional neural network, which was trained and evaluated on PSG recordings from 15,660 participants of 16 clinical studies. It provides accurate segmentations across a wide range of patient cohorts and PSG protocols not considered when building the system. U-Sleep works for arbitrary combinations of typical EEG and EOG channels, and its special deep learning architecture can label sleep stages at shorter intervals than the typical 30 s periods used during training. We show that these labels can provide additional diagnostic information and lead to new ways of analyzing sleep. U-Sleep performs on par with state-of-the-art automatic sleep staging systems on multiple clinical datasets, even if the other systems were built specifically for the particular data. A comparison with consensus-scores from a previously unseen clinic shows that U-Sleep performs as accurately as the best of the human experts. U-Sleep can support the sleep staging workflow of medical experts, which decreases healthcare costs, and can provide highly accurate segmentations when human expertize is lacking.

B-waves are present in patients without intracranial pressure disturbances.

Intracranial pressure (ICP) B-waves are defined as short, repeating elevations of ICP of up to 50 mmHg with a frequency of 0.5-2 waves/min. The presence of B-waves in overnight recordings is regarded as a pathological phenomenon. However, the physiology of B-waves is still not fully understood and studies with transcranial Doppler, as a surrogate marker for ICP, have suggested that B-waves could be a normal physiological phenomenon. We present four patients without known structural neurological disease other than a coincidentally found unruptured intracranial aneurysm. One of the patients had experienced well-controlled epilepsy for several years, but was included because ICP under these conditions is unlikely to be abnormal. Following informed consent, all four patients had a telemetric ICP probe implanted during a prophylactic operation with closure of the aneurysm. They underwent overnight ICP monitoring with simultaneous polysomnography (PSG) sleep studies at 8 weeks after the operation. These patients exhibited nocturnal B-waves, but did not have major structural brain lesions. Their ICP values were within the normal range. Nocturnal B-waves occurred in close association with sleep-disordered breathing (SDB) in rapid eye movement (REM) and non-REM sleep stages. SDB during REM sleep was associated with ramp-type B-waves; SDB during non-REM sleep was associated with the sinusoidal type of B-wave. We propose that B-waves are a physiological phenomenon associated with SDB and that the mechanical changes during respiration could have an essential and previously unrecognised role in the generation of B-waves.

Comparison of computerized methods for rapid eye movement sleep without atonia detection.

Rapid eye movement (REM) sleep without atonia detection is a prerequisite for diagnosis of REM sleep behavior disorder (RBD). As the visual gold standard method is time-consuming and subjective, several automated methods have been proposed. This study aims to compare their performances: The REM atonia index (RAI), the supra-threshold-REM-activity metric, the Frandsen index, the short/long muscle activity indices, and the Kempfner index algorithms were applied to 27 healthy control participants (C), 25 patients with Parkinson's disease (PD) without RBD (PD-RBD), 29 patients with PD and RBD (PD + RBD), 29 idiopathic patients with RBD, and 36 patients with periodic limb movement disorder (PLMD). The indices were calculated in various configurations: (1) considering all muscle activities; (2) excluding the ones related to arousals; (3) excluding the ones during apnea events; (4) excluding the ones before and after apnea events; (5) combining configurations 2 and 3; and (6) combining configurations 2 and 4. For each of these configurations, the discrimination capability of the indices was tested for the following comparisons: (1) (C, PD-RBD, PLMD) vs (PD + RBD, RBD); (2) C vs RBD; (3) PLMD vs RBD; (4) C vs PD-RBD; (5) C vs PLMD; (6) PD-RBD vs PD + RBD; and (7) C vs PLMD vs RBD. Results showed varying methods' performances across the different configurations and comparisons, making it impossible to identify the optimal method and suggesting the need of further improvements. Nevertheless, RAI seems the most sensible one for RBD detection. Moreover, apnea and arousal-related movements seem not to influence the algorithms' performances in patients' classification.

Novel method for evaluation of eye movements in patients with narcolepsy.

BACKGROUND: Narcolepsy causes abnormalities in the control of wake-sleep, non-rapid-eye-movement (non-REM) sleep and REM sleep, which includes specific eye movements (EMs). In this study, we aim to evaluate EM characteristics in narcolepsy as compared to controls using an automated detector. METHODS: We developed a data-driven method to detect EMs during sleep based on two EOG signals recorded as part of a polysomnography (PSG). The method was optimized using the manually scored hypnograms from 36 control subjects. The detector was applied on a clinical sample with subjects suspected for central hypersomnias. Based on PSG, multiple sleep latency test and cerebrospinal fluid hypocretin-1 measures, they were divided into clinical controls (N = 20), narcolepsy type 2 (NT2, N = 19), and narcolepsy type 1 (NT1, N = 28). We investigated the distribution of EMs across sleep stages and cycles. RESULTS: NT1 patients had significantly less EMs during wake, N1, and N2 sleep and more EMs during REM sleep compared to clinical controls, and significantly less EMs during wake and N1 sleep compared to NT2 patients. Furthermore, NT1 patients showed less EMs during NREM sleep in the first sleep cycle and more EMs during NREM sleep in the second sleep cycle compared to clinical controls and NT2 patients. CONCLUSIONS: NT1 patients show an altered distribution of EMs across sleep stages and cycles compared to NT2 patients and clinical controls, suggesting that EMs are directly or indirectly controlled by the hypocretinergic system. A data-driven EM detector may contribute to the evaluation of narcolepsy and other disorders involving the control of EMs.

Analysis of automated quantification of motor activity in REM sleep behaviour disorder.

Rapid eye movement (REM) sleep behaviour disorder (RBD) is characterized by dream enactment and REM sleep without atonia. Atonia is evaluated on the basis of visual criteria, but there is a need for more objective, quantitative measurements. We aimed to define and optimize a method for establishing baseline and all other parameters in automatic quantifying submental motor activity during REM sleep. We analysed the electromyographic activity of the submental muscle in polysomnographs of 29 patients with idiopathic RBD (iRBD), 29 controls and 43 Parkinson's (PD) patients. Six adjustable parameters for motor activity were defined. Motor activity was detected and quantified automatically. The optimal parameters for separating RBD patients from controls were investigated by identifying the greatest area under the receiver operating curve from a total of 648 possible combinations. The optimal parameters were validated on PD patients. Automatic baseline estimation improved characterization of atonia during REM sleep, as it eliminates inter/intra-observer variability and can be standardized across diagnostic centres. We found an optimized method for quantifying motor activity during REM sleep. The method was stable and can be used to differentiate RBD from controls and to quantify motor activity during REM sleep in patients with neurodegeneration. No control had more than 30% of REM sleep with increased motor activity; patients with known RBD had as low activity as 4.5%. We developed and applied a sensitive, quantitative, automatic algorithm to evaluate loss of atonia in RBD patients.

EEG recordings as a source for the detection of IRBD.

The purpose of this pilot study was to develop a supportive algorithm for the detection of idiopathic Rapid Eye-Movement (REM) sleep Behaviour Disorder (iRBD) from EEG recordings. iRBD is defined as REM sleep without atonia with no current sign of neurodegenerative disease, and is one of the earliest known biomarkers of Parkinson's Disease (PD). It is currently diagnosed by polysomnography (PSG), primarily based on EMG recordings during REM sleep. The algorithm was developed using data collected from 42 control subjects and 34 iRBD subjects. A feature was developed to represent high amplitude contents of the EEG and a semi-automatic signal reduction method was introduced. The reduced feature set was used for a subject-based classification. With a subject specific re-scaling of the feature set and the use of an outlier detection classifier the algorithm reached an accuracy of 0.78. The result shows that EEG recordings contain valid information for a supportive algorithm for the detection of iRBD. Further investigation could lead to promising application of EEG recordings as a supportive source for the detection of iRBD.

Novel method for detection of Sleep Apnoea using respiration signals.

Polysomnography (PSG) studies are considered the "gold standard" for the diagnosis of Sleep Apnoea (SA). Identifying cessations of breathing from long-lasting PSG recordings manually is a labour-intensive and time-consuming task for sleep specialist, associated with inter-scorer variability. In this study a simplified, semi-automatic, three-channel method for detection of SA patients is proposed in order to increase analysis reliability and diagnostic accuracy in the clinic. The method is based on characteristic features, such as respiration stoppages pr. hour and the total number of oxygen desaturations > 3%, extracted from the thorax and abdomen respiration effort belts, and the oxyhemoglobin saturation (SaO2), fed to an Elastic Net classifier and validated according to American Academy of Sleep Medicine (AASM) using the patients' AHI value. The method was applied to 109 patient recordings and resulted in a very high SA classification with accuracy of 97.9%. The proposed method reduce the time spent on manual analysis of respiration stoppages and the inter- and intra-scorer variability, and may serve as an alternative screening method for SA.