Merel Marietje van Gilst

Objective: Wrist-worn photoplethysmography (PPG) enables scalable, long-term unobtrusive sleep monitoring through the expression of sympathetic and parasympathetic activity in heart rate variability. However, interindividual differences in the sympatho-vagal balance may inherently limited general PPG-based sleep staging models. This study investigates whether learning individual autonomic representations through model personalization can improve PPG based automated sleep staging performance.- Approach: Concurrent wrist-worn PPG and wearable electroencephalography (EEG) were collected during home monitoring for up to seven nights in a heterogeneous sleep-disordered population (n=59). Personalization was performed through finetuning (i.e., partial retraining) a general PPG-based model by coupling the subject-specific PPG data with the wearable EEG stage classifications. Performance of the general and personalized models were compared on PPG acquired during a gold-standard clinical polysomnography, testing their agreement on 4-stage classification (W/N1+N2/N3/REM) with the manual scoring. Main Result: Overall performance increased in 82.5% of the subjects, with significantly improved performance reached when personalizing the model on three or more training nights. Performance increased with personalization on additional training nights for each stage: wake (β=.005, p<.001), N1+N2 (β=.003, p<.001), N3 (β=.004, p<.001), and REM (β=.005, p<.001). Effects were strongest for younger individuals (β=.009, p<.001) and patients with insomnia (β=.011, p<.001). Personalization greatly impacted the derived sleep macrostructural sleep parameters, with considerable improvement in N3 sleep classification, and in capturing REM fragmentation. Significance: Personalization can overcome one-size-fits-all limitations of a general model and should be considered for PPG-based sleep staging when an altered autonomic modulation is expected that deviates from the general model's global representation.

Automatic sleep staging from single-channel electroencephalography (EEG) using artificial intelligence (AI) is emerging as an alternative to costly and time-consuming manual scoring using multi-channel polysomnography. However, current AI methods, mainly deep learning models such as convolutional neural network (CNN) and long short-term memory (LSTM), struggle to detect the N1 sleep stage, which is challenging due to its rarity and ambiguous nature compared to other stages. Here we propose SSC-SleepNet, an automatic sleep staging algorithm aimed at improving the learning of N1 sleep. SSC-SleepNet employs a pseudo-Siamese neural network architecture owing to its capability in one- or few-shot learning with contrastive loss. Which we selected due to its strong capability in one- or few-shot learning with a contrastive loss function. SSC-SleepNet consists of two branches of neural networks: a squeeze-and-excitation residual network branch and a CNN-LSTM branch. These two branches are used to generate latent features of the EEG epoch. The adaptive loss function of SSC-SleepNet uses a weighing factor to combine weighted cross-entropy loss and focal loss to specifically address the class imbalance issue inherent in sleep staging. The proposed new loss function dynamically assigns a higher penalty to misclassified N1 sleep stages, which can improve the model's learning capability for this minority class. Four datasets were used for sleep staging experiments. In the Sleep-EDF-SC, Sleep-EDF-X, Sleep Heart Health Study, and Haaglanden Medisch Centrum datasets, SSC-SleepNet achieved macro F1-scores of 84.5%, 89.6%, 89.5%, and 85.4% for all sleep stages, and N1 sleep stage F1-scores of 60.2%, 58.3%, 57.8%, and 55.2%, respectively. Our proposed deep learning model outperformed the most existing models in automatic sleep staging using single-channel EEG signals. In particular, N1 detection performance has been markedly improved compared to the state-of-art models.

Gold-standard sleep scoring is based on epoch-based assignment of sleep stages based on a combination of EEG, EOG and EMG signals. However, a polysomnographic recording consists of many other signals that could be used for sleep staging, including cardio-respiratory modalities. Leveraging this signal variety would offer important advantages, for example increasing reliability, resilience to signal loss, and application to long-term non-obtrusive recordings. We developed a deep generative model for automatic sleep staging from a plurality of sensors and any -arbitrary- combination thereof. We trained a score-based diffusion model using a dataset of 1947 expert-labelled overnight recordings with 36 different signals, and achieved zero-shot inference on any sensor set by leveraging a novel Bayesian factorization of the score function across the sensors. On single-channel EEG, the model reaches the performance limit in terms of polysomnography inter-rater agreement (5- class accuracy 85.6%, Cohen's kappa 0.791). Moreover, the method offers full flexibility to use any sensor set, for example finger photoplethysmography, nasal flow and thoracic respiratory movements, (5-class accuracy 79.0%, Cohen's kappa of 0.697), or even derivations very unconventional for sleep staging, such as tibialis and sternocleidomastoid EMG (5-class accuracy 71.0%, kappa 0.575). Additionally, we propose a novel interpretability metric in terms of information gain per sensor and show this is linearly correlated with classification performance. Finally, our model allows for post- hoc addition of entirely new sensor modalities by merely training a score estimator on the novel input instead of having to retrain from scratch on all inputs.

Rapid‐eye‐movement (REM) sleep behaviour disorder (RBD) is a primary sleep disorder strongly associated with Parkinson's disease. Assessing sleep structure in RBD is important for understanding the underlying pathophysiology and developing diagnostic methods. However, the performance of automated sleep stage classification (ASSC) models is considered suboptimal in RBD, for both models utilising neurological signals (“ExG”: EEG, EOG, and chin EMG) and heart rate variability combined with body movements (HRVm). Here, we explore this underperformance through the categorical representation of sleep macrostructure (i.e., hypnogram) and a representation that leverages the underlying probability distribution of ASSCs (i.e., hypnodensity). By comparing the RBD population ( n = 36) to a sex‐ and age‐matched group of OSA patients chosen for their anticipated similarly decreased sleep stability, we confirm lower 4‐stage classification performance in both ExG‐based ASSC (RBD: κ = 0.74, OSA: κ = 0.80) and HRVm‐based ASSC (RBD: κ = 0.50, OSA: κ = 0.63). Stages showing lower agreement in RBD, namely, N1 + N2 and REM sleep, exhibited elevated ambiguity in the hypnodensity, indicating more ambiguous classification distributions. Limited differences in bout durations between RBD and OSA suggested sleep instability is not necessarily driving lower agreement in RBD. However, stage transitions in OSA showed more abrupt changes in the underlying probability distribution, while RBD transitions had a more continuous profile, possibly complicating classification. Although both ExG‐based and HRVm‐based automated sleep staging in RBD remain challenging, hypnodensity analysis is informative for the characterisation of (RBD) sleep and can capture potential drivers of classification disagreement.

Total sleep time (TST) misperception has been reported in obstructive sleep apnea (OSA). However, previous findings on predictors were inconsistent and predominantly relied on single-night polysomnography, which may alter patients’ sleep perception. We leveraged advances in wearable sleep staging to investigate predictors of TST misperception in OSA over multiple nights in the home environment. The study included 141 patients with OSA, 75 without insomnia symptoms (OSA group), and 66 with insomnia symptoms (OSA-I group). Objective TST was measured using a previously validated wrist-worn photoplethysmography and accelerometry device. Self-reported TST was assessed using a digital sleep diary. TST misperception was quantified with the misperception index (MI), calculated as (objective − self-reported TST)/objective TST. MI values differed significantly between the OSA (median = −0.02, IQR = [−0.06, 0.02]) and the OSA-I group (0.05, [−0.02, 0.13], p < 0.001). Multilevel modeling revealed that the presence of insomnia symptoms (β = 0.070, p < 0.001) and lower daily reported sleep quality (β = −0.229, p < 0.001) were predictive of higher MI (TST underestimation), while a higher apnea–hypopnea index (AHI) was predictive of lower MI (TST overestimation; β = −0.001, p = 0.006). Thus, insomnia symptoms and AHI are associated with TST misperception in OSA patients, but in opposite directions. This association extends over multiple nights in the home environment.

Objective. Sleep staging based on full polysomnography is the gold standard in the diagnosis of many sleep disorders. It is however costly, complex, and obtrusive due to the use of multiple electrodes. Automatic sleep staging based on single-channel electro-oculography (EOG) is a promising alternative, requiring fewer electrodes which could be self-applied below the hairline. EOG sleep staging algorithms are however yet to be validated in clinical populations with sleep disorders. Approach. We utilized the SOMNIA dataset, comprising 774 recordings from subjects with various sleep disorders, including insomnia, sleep-disordered breathing, hypersomnolence, circadian rhythm disorders, parasomnias, and movement disorders. The recordings were divided into train (574), validation (100), and test (100) groups. We trained a neural network that integrated transformers within a U-Net backbone. This design facilitated learning of arbitrary-distance temporal relationships within and between the EOG and hypnogram. Main results. For 5-class sleep staging, we achieved median accuracies of 85.0% and 85.2% and Cohen’s kappas of 0.781 and 0.796 for left and right EOG, respectively. The performance using the right EOG was significantly better than using the left EOG, possibly because in the recommended AASM setup, this electrode is located closer to the scalp. The proposed model is robust to the presence of a variety of sleep disorders, displaying no significant difference in performance for subjects with a certain sleep disorder compared to those without. Significance. The results show that accurate sleep staging using single-channel EOG can be done reliably for subjects with a variety of sleep disorders.

Objective: wearable sensor technology has progressed significantly in the last decade, but its clinical usability for the assessment of obstructive sleep apnea (OSA) is limited by the lack of large and representative datasets simultaneously acquired with polysomnography (PSG). The objective of this study was to explore the use of cardiorespiratory signals commonly available in standard PSGs which can be easily measured with wearable sensors, to estimate the severity of OSA. Methods: an artificial neural network was developed for detecting sleep disordered breathing events using electrocardiogram (ECG) and respiratory effort. The network was combined with a previously developed cardiorespiratory sleep staging algorithm and evaluated in terms of sleep staging classification performance, apnea-hypopnea index (AHI) estimation, and OSA severity estimation against PSG on a large cohort of 653 participants with a wide range of OSA severity. Results: four-class sleep staging achieved a κ of 0.69 with PSG, distinguishing wake, combined N1-N2, N3 and REM. AHI estimation achieved an intraclass correlation coefficient of 0.91, and high diagnostic performance for different OSA severity thresholds. Conclusions: this study highlights the potential of using cardiorespiratory signals to estimate OSA severity, even without the need for airflow or oxygen saturation (SpO2), traditionally used for assessing OSA. Significance: while further research is required to translate these findings to practical and unobtrusive sensors, this study demonstrates how existing, large datasets can serve as a foundation for wearable systems for OSA monitoring. Ultimately, this approach could enable long-term assessment of sleep disordered breathing, facilitating new avenues for clinical research in this field.

Summary: Non‐rapid eye movement parasomnia disorders, also called disorders of arousal, are characterized by abnormal nocturnal behaviours, such as confusional arousals or sleep walking. Their pathophysiology is not yet fully understood, and objective diagnostic criteria are lacking. It is known, however, that behavioural episodes occur mostly in the beginning of the night, after an increase in slow‐wave activity during slow‐wave sleep. A better understanding of the prospect of such episodes may lead to new insights in the underlying mechanisms and eventually facilitate objective diagnosis. We investigated temporal dynamics of transitions from slow‐wave sleep of 52 patients and 79 controls. Within the patient group, behavioural and non‐behavioural N3 awakenings were distinguished. Patients showed a higher probability to wake up after an N3 bout ended than controls, and this probability increased with N3 bout duration. Bouts longer than 15 min resulted in an awakening in 73% and 34% of the time in patients and controls, respectively. Behavioural episodes reduced over sleep cycles due to a reduction in N3 sleep and a reducing ratio between behavioural and non‐behavioural awakenings. In the first two cycles, N3 bouts prior to non‐behavioural awakenings were significantly shorter than N3 bouts advancing behavioural awakenings in patients, and N3 awakenings in controls. Our findings provide insights in the timing and prospect of both behavioural and non‐behavioural awakenings from N3, which may result in prediction and potentially prevention of behavioural episodes. This work, moreover, leads to a more complete characterization of a prototypical hypnogram of parasomnias, which could facilitate diagnosis.

Purpose Comorbid insomnia often occurs in patients with obstructive sleep apnea (OSA), referred to as COMISA. Cortical arousals manifest as a common feature in both OSA and insomnia, often accompanied by elevated heart rate (HR). Our objective was to evaluate the heart rate response to nocturnal cortical arousals in patients with COMISA and patients with OSA alone. Methods We analyzed data from patients with COMISA and from patients with OSA matched for apnea-hypopnea index. Sleep staging and analysis of respiratory events and cortical arousals were performed using the Philips Somnolyzer automatic scoring system. Beat-by-beat HR was analyzed from the onset of the cortical arousal to 30 heartbeats afterwards. HR responses were divided into peak and recovery phases. Cortical arousals were separately evaluated according to subtype (related to respiratory events and spontaneous) and duration (3–6 s, 6–10 s, 10–15 s). Results A total of 72 patients with COMISA and 72 patients with OSA were included in this study. There were no overall group differences in the number of cortical arousals with and without autonomic activation. No significant differences were found for spontaneous cortical arousals. The OSA group had more cortical arousals related to respiratory events (21.0 [14.8–30.0] vs 16.0 [9.0–27.0], p = 0.016). However, the COMISA group had longer cortical arousals (7.2 [6.4–7.8] vs 6.7 [6.2–7.7] s, p = 0.024) and the HR recovery phase was prolonged (52.5 [30.8–82.5] vs 40.0 [21.8–55.5] beats/min, p = 0.017). Both the peak and the recovery phase for longer cortical arousals with a duration of 10–15 s were significantly higher in patients with COMISA compared to patients with OSA (47.0 [27.0–97.5] vs 34.0 [21.0–71.0] beats/min, p = 0.032 and 87.0 [47.0–132.0] vs 71.0 [43.0–103.5] beats/min, p = 0.049, respectively). Conclusions The HR recovery phase after cortical arousals related to respiratory events is prolonged in patients with COMISA compared to patients with OSA alone. This response could be indicative of the insomnia component in COMISA.

Sleep staging is the process by which an overnight polysomnographic measurement is segmented into epochs of 30 seconds, each of which is annotated as belonging to one of five discrete sleep stages. The resulting scoring is graphically depicted as a hypnogram, and several overnight sleep statistics are derived, such as total sleep time and sleep onset latency. Gold standard sleep staging as performed by human technicians is time-consuming, costly, and comes with imperfect inter-scorer agreement, which also results in inter-scorer disagreement about the overnight statistics. Deep learning algorithms have shown promise in automating sleep scoring, but struggle to model inter-scorer disagreement in sleep statistics. To that end, we introduce a novel technique using conditional generative models based on Normalizing Flows that permits the modeling of the inter-rater disagreement of overnight sleep statistics, termed U-Flow. We compare U-Flow to other automatic scoring methods on a hold-out test set of 70 subjects, each scored by six independent scorers. The proposed method achieves similar sleep staging performance in terms of accuracy and Cohen's kappa on the majority-voted hypnograms. At the same time, U-Flow outperforms the other methods in terms of modeling the inter-rater disagreement of overnight sleep statistics. The consequences of inter-rater disagreement about overnight sleep statistics may be great, and the disagreement potentially carries diagnostic and scientifically relevant information about sleep structure. U-Flow is able to model this disagreement efficiently and can support further investigations into the impact inter-rater disagreement has on sleep medicine and basic sleep research.