Sleep disorders are diagnosed in sleep laboratories by polysomnography, a multi-parameter examination that monitors biological signals during sleep. The subsequent evaluation of the obtained records is very time-consuming. The goal of this study was to create an automatic system for evaluation of the airflow and SpO${}_2$ channels of polysomnography records, through the use of machine learning techniques and a large database, for apnea and desaturation detection (which is unusual in other studies). To that end, a convolutional neural network (CNN) was designed using hyperparameter optimization. It was then trained and tested for apnea and desaturation. The proposed CNN was compared with the commonly used k-nearest neighbors (k-NN) method. The classifiers were designed based on nasal airflow and blood oxygen saturation signals. The final neural network accuracy for apnea detection reached 84%, and that for desaturation detection was 74%, while the k-NN classifier reached accuracies of 83% and 64% for apnea detection and desaturation detection, respectively.

Summary: Insomnia is a prevalent and disabling condition whose treatment is not always effective. This pilot study explores the feasibility and effects of closed‐loop auditory stimulation (CLAS) as a potential non‐invasive intervention to improve sleep, its subjective quality, and memory consolidation in patients with insomnia. A total of 27 patients with chronic insomnia underwent a crossover, sham‐controlled study with 2 nights of either CLAS or sham stimulation. Polysomnography was used to record sleep parameters, while questionnaires and a word‐pair memory task were administered to assess subjective sleep quality and memory consolidation. The initial analyses included 17 patients who completed the study, met the inclusion criteria, and received CLAS. From those, 10 (58%) received only a small number of stimuli. In the remaining seven (41%) patients with sufficient CLAS, we evaluated the acute and whole‐night effect on sleep. CLAS led to a significant immediate increase in slow oscillation (0.5–1 Hz) amplitude and activity, and reduced delta (1–4 Hz) and sigma/sleep spindle (12–15 Hz) activity during slow‐wave sleep across the whole night. All these fundamental sleep rhythms are implicated in sleep‐dependent memory consolidation. Yet, CLAS did not change sleep‐dependent memory consolidation or sleep macrostructure characteristics, number of arousals, or subjective perception of sleep quality. Results showed CLAS to be feasible in patients with insomnia. However, a high variance in the efficacy of our automated stimulation approach suggests that further research is needed to optimise stimulation protocols to better unlock potential CLAS benefits for sleep structure and subjective sleep quality in such clinical settings.

Slow-wave synchronous acoustic stimulation is a promising research and therapeutic tool. It is essential to clearly understand the principles of the synchronization methods, to know their performances and limitations, and, most importantly, to have a clear picture of the effect of stimulation on slow-wave activity (SWA). This paper covers the mentioned and currently missing parts of knowledge that are essential for the appropriate development of the method itself and future applications. Artificially streamed real sleep EEG data were used to quantitatively compare the two currently used real-time methods: the phase-locking loop (PLL) and the fixed-step stimulus in our own implementation. The fixed-step stimulation method was concluded to be more reliable and practically applicable compared to the PLL method. The sleep experiment with chronic insomnia patients in our sleep laboratory was analyzed in order to precisely characterize the effect of sound stimulation during deep sleep. We found that there is a significant phase synchronization of delta waves, which were shown to be the most sensitive metric of the effect of acoustic stimulation compared to commonly used averaged signal and power analyses. This finding may change the understanding of the effect and function of the SWA stimulation described in the literature.

Functional magnetic resonance imaging (fMRI) techniques and electroencephalography (EEG) were used to investigate sleep with a focus on impaired arousal mechanisms in disorders of arousal (DOAs). With a prevalence of 2–4% in adults, DOAs are significant disorders that are currently gaining attention among physicians. The paper describes a simultaneous EEG and fMRI experiment conducted in adult individuals with DOAs ($n=10$). Both EEG and fMRI data were validated by reproducing well established EEG and fMRI associations. A method for identification of both brain functional areas and EEG rhythms associated with DOAs in shallow sleep was designed. Significant differences between patients and controls were found in delta, theta, and alpha bands during awakening epochs. General linear models of the blood-oxygen-level-dependent signal have shown the secondary visual cortex and dorsal posterior cingulate cortex to be associated with alpha spectral power fluctuations, and the precuneus with delta spectral power fluctuations, specifically in patients and not in controls. Future EEG–fMRI sleep studies should also consider subject comfort as an important aspect in the experimental design.

The electroencephalogram (EEG) reflecting brain activity may be characterised through brief periods of stable neural activity patterns that recur over time and are referred to as microstates. Microstates are related to a range of cognitive processes, and their analysis has become an increasingly popular tool for studying human brain function. While microstates have been extensively studied in humans, their presence and characteristics in animal models have yet to be as thoroughly investigated. This study aims to address this gap by detecting and characterising microstates in EEGs of rats collected using a superficial electrode system corresponding to homological areas of the human 10–20 system. Specifically, we demonstrate the presence of microstates in rats’ EEGs, i.e., those that may be captured by the same metrics as in humans. We identified five microstate EEG maps in rats, explaining 71% of the variance in our dataset ( N = 30). The explained variance, mean temporal coverage values (0.2), and average duration (0.26 s) are comparable to the human-derived EEG microstates. Via a source localisation technique, the cingulate cortex, precuneus, and insula were found to be associated with the microstates’ temporal dynamics. Among the microstates that showed a broadband character, we also found those that showed an association with the theta and alpha bands. These findings have important implications for the use of microstates as a preclinical tool for investigating brain functions, detecting new biomarkers of brain diseases, and translating this knowledge to humans.