Sleep stage scoring using the neural network model: comparison between visual and automatic analysis in normal subjects and patients.

To investigate the pathophysiology of narcoleptic patients' sleep in detail, we analysed and compared the whole-night polysomnograms of narcoleptic patients and normal human subjects. Eight drug-naive narcoleptic patients and eight age-matched normal volunteers underwent polysomnography (PSG) on two consecutive nights. In addition to conventional visual scoring of the polysomnograms, rapid eye movement (REM)-density and electroencephalograph (EEG) power spectra analyses were also performed. Sleep onset REM periods and fragmented nocturnal sleep were observed as expected in our narcoleptic patients. In the narcoleptic patients, REM period duration across the night did not show the significant increasing trend that is usually observed in normal subjects. In all narcoleptic patient REM periods, eye movement densities were significantly increased. The power spectra of narcoleptic REM sleep significantly increased between 0.3 and 0.9 Hz and decreased between 1.0 and 5.4 Hz. Further analysis revealed that non-rapid eye movement (NREM) period duration and the declining trend of delta power density in the narcoleptic patients were not significantly different from the normal subjects. Compared with normal subjects, the power spectra of narcoleptic NREM sleep increased in the 1.0-1.4 Hz and 11.0-11.9 Hz frequency bands, and decreased in a 24.0-26.9 Hz frequency band. Thus, increased EEG delta and decreased beta power densities were commonly observed in both the NREM and REM sleep of the narcoleptic patients, although the decrease in beta power during REM sleep was not statistically significant. Our visual analysis revealed fragmented nocturnal sleep and increased phasic REM components in the narcoleptic patients, which suggest the disturbance of sleep maintenance mechanism(s) and excessive effects of the mechanism(s) underlying eye movement activities during REM sleep in narcolepsy. Spectral analysis revealed significant increases in delta components and decreases in beta components, which suggest decreased activity in central arousal mechanisms. These characteristics lead us to hypothesize that two countervailing mechanisms underlie narcoleptic sleep pathology.

Feeling awake although sleep recordings indicate clear-cut sleep sometimes occurs in good sleepers and to an extreme degree in patients with so-called paradoxical insomnia. It is unknown what underlies sleep misperception, as standard polysomnographic (PSG) parameters are often normal in these cases. Here we asked whether regional changes in brain activity could account for the mismatch between objective and subjective total sleep times (TST). To set cutoffs and define the norm, we first evaluated sleep perception in a population-based sample, consisting of 2,092 individuals who underwent a full PSG at home and estimated TST the next day. We then compared participants with a low mismatch (normoestimators, n = 1,147, ±0.5 SD of mean) with those who severely underestimated (n = 52, <2.5th percentile) or overestimated TST (n = 53, >97.5th percentile). Compared with normoestimators, underestimators displayed higher electroencephalographic (EEG) activation (beta/delta power ratio) in both rapid eye movement (REM) and non-rapid eye movement (NREM) sleep, while overestimators showed lower EEG activation (significant in REM sleep). To spatially map these changes, we performed a second experiment, in which 24 healthy subjects and 10 insomnia patients underwent high-density sleep EEG recordings. Similarly to underestimators, patients displayed increased EEG activation during NREM sleep, which we localized to central-posterior brain areas. Our results indicate that a relative shift from low- to high-frequency spectral power in central-posterior brain regions, not readily apparent in conventional PSG parameters, is associated with underestimation of sleep duration. This challenges the concept of sleep misperception, and suggests that instead of misperceiving sleep, insomnia patients may correctly perceive subtle shifts toward wake-like brain activity.

To study the effects of pulmonary denervation on breathing during sleep, sleep studies were conducted on seven heart-lung transplant recipients (H-LT) and a comparable number of sex-matched normal subjects of similar age. Four of the H-LT patients had a restrictive pattern on spirometry. The time since transplantation ranged from 45 to 1,102 days. There were no significant differences between the groups with respect to total sleep time or distribution of sleep stages. There were no significant differences between the H-LT recipients and normal subjects with respect to baseline awake oxyhemoglobin saturation (SaO2) or the nadirs of SaO2 during REM and non-REM sleep, the absolute number and frequency (number per hour of sleep) of apneas, hypopneas, desaturation events, both over the whole night of study or separately during non-REM and REM sleep. Across wakefulness and all sleep stages, the H-LT patients tended to have shorter total respiratory cycle times (Ttot) (p = 0.052) and more rapid breathing frequency (F) than the normal subjects. This was associated with significantly shorter inspiratory times (Tl) (p less than 0.001) and smaller duty cycles (Tl/Ttot) (p less than 0.005) in the H-LT recipients. During non-REM and REM sleep, F tended to be higher in the H-LT recipients with pulmonary restriction than in the nonrestricted patients. There were no significant differences between the H-LT recipients and the normal subjects with regard to the periodicity of breathing, either in terms of timing parameters or breath amplitude.(ABSTRACT TRUNCATED AT 250 WORDS)

On-line automatic sleep scoring system involving Bayesian filtering

Sleep state instability is a potential mechanism of central apnea/hypopnea during non-rapid eye movement (NREM) sleep. To investigate this postulate, we induced brief arousals by delivering transient (0.5 second) auditory stimuli during stable NREM sleep in eight normal subjects. Arousal was determined according to American Sleep Disorders Association (ASDA) criteria. A total of 96 trials were conducted; 59 resulted in cortical arousal and 37 did not result in arousal. In trials associated with arousal, minute ventilation (VE) increased from 5.1 +/- 1.24 minutes to 7.5 +/- 2.24 minutes on the first posttone breath (p = 0.001). However, no subsequent hypopnea or apnea occurred as VE decreased gradually to 4.8 +/- 1.5 l/minute (p > 0.05) on the fifth posttone breath. Trials without arousal did not result in hyperpnea on the first breath nor subsequent hypopnea. We conclude that 1) auditory stimulation resulted in transient hyperpnea only if associated with cortical arousal; 2) hypopnea or apnea did not occur following arousal-induced hyperpnea in normal subjects; 3) interaction with fluctuating chemical stimuli or upper airway resistance may be required for arousals to cause sleep-disordered breathing.

The present study aims to explore the effectiveness of the automatic pronunciation scoring system of a newly released AI chatbot, ‘AI Pengtalk’. This English educational application, designed by EBSe and ETRI, was distributed to public elementary schools nationwide free of charge in April, 2021. This study investigates 6th graders’ recognition of AI Pengtalk and the effects of four subscales of the pronunciation score given by AI Pengtalk: intonation; segment accuracy; utterance speed; stress and rhythm. It also examines the correlation between holistic pronunciation scores and students’ English achievement test scores. The results show that 6th graders have quite positive attitudes to AI Pengtalk, and suggest that the instant feedback of AI Pengtalk’s automatic scoring system appears to have little effect. It is also shown that the subscale stress and rhythm significantly influences the corresponding holistic pronunciation score and that the correlation between pronunciation and English achievement seems to be different according to the consisting phonemes of the specific sentences. The present study suggests that elementary students’ pronunciation can be improved by the automatic scoring system of AI chatbots. Further research is recommended with the comprehensive data collection to draw more pedagogical implications.

Conscious experiences and high-density EEG patterns predicting subjective sleep depth

studies using a variety of methods (transcranial Doppler ultrasonography,133Xe inhalation, and single-photon emission computerized tomography) have shown a reduction (∼10–20%) in cerebral blood flow (CBF) during non-rapid-eye-movement (NREM) sleep and an increase (∼20–35%) during rapid-eye-

Introduction Sleep is a fundamental necessity for health and is commonly disrupted in the perioperative period. Technological improvements leveraging dry electroencephalographic (EEG) sensors have opened the door for large-scale quantitative assessments of sleep in relation to perioperative outcomes. Methods Patients utilized the Dreem (Rhythm, New York USA), a wireless EEG headband, to acquire their own preoperative nocturnal sleep records at home. Following cardiac surgery, postoperative recordings were obtained with staff assistance until postoperative night 7. Sleep records were scored as rapid eye movement (REM) and non-rapid eye movement (NREM) stages N1-N3, using modified American Academy of Sleep Medicine guidelines. Results Of 100 patients enrolled for perioperative sleep recordings, 74 patients provided 132 preoperative records; 80% were scorable with a median total sleep time (TST) of 209.8 minutes. TST was distributed as 8.3% N1, 70.6% N2, 2.1% N3 and 19% REM, consistent with expected sleep structure in geriatric populations. EEG markers for staging sleep were evaluated in the scorable records: 92% with sleep spindles, 98% with K-complexes, 69% with slow waves, 92% with sawtooth waves, and 80% with rapid eye movements. Among 26 patients with multiple preoperative sleep recordings, no significant within-subject differences in sleep structure were observed (all p &amp;gt; 0.05, paired Wilcoxon sign-rank test). 270 postoperative nocturnal sleep recordings were obtained from 83 patients, 70% of which were scorable. TST in scorable postoperative records was distributed as 14.9% N1, 78.6% N2, 0.9% N3 and 5.6% REM. Durations of REM and N3 sleep were significantly reduced in postoperative (POD 1-4) overnight recordings compared to preoperative measurements (Skillings–Mack test, p &amp;lt; 0.001 and p = 0.02 for REM and N3, respectively). Conclusion Wireless EEG devices enhance the feasibility of assaying perioperative sleep. A single night of unattended, ambulatory sleep monitoring is sufficient to establish a preoperative baseline. Multiple preoperative and postoperative sleep studies were tolerated by patients, which showed reductions of N3 and REM sleep in the early postoperative period. This study demonstrates the feasibility of using the Dreem for monitoring sleep macro- and microstructural EEG elements in the perioperative setting. Support (if any):

Introduction Polysomnography, the gold standard for diagnosing obstructive sleep apnea and other sleep disorders, requires manual scoring by trained sleep technologists or physicians to analyze its comprehensive physiological data. However, manual scoring is cost-intensive, prone to inconsistencies among scorers, and often causes delays in diagnosis and treatment initiation, potentially compromising patient outcomes. The purpose of this study is to validate the performance of artificial intelligence (AI)-based automatic polysomnography scoring system (SOMNUM) developed to overcome these limitations using a cross-sectional experimental design on a representative subjects of data recorded in the 3 sleep laboratories of university hospitals. Methods This study utilized a retrospective dataset of 400 polysomnography (PSG) recordings from three university hospitals: Soonchunhyang University Bucheon Hospital, Ajou University Hospital, and Chungnam University Hospital. Forty-eight full-night PSG studies were randomly selected from the dataset for the performance evaluation of SOMNUM, an AI-based automatic sleep scoring system. The performance of SOMNUM was assessed by comparing its scoring results with those of three human experts for four scoring functions: sleep staging, arousals, respiratory events, and periodic limb movements (PLMs). Performance metrics included positive percentage agreement, negative percentage agreement, and overall percentage agreement, calculated using R statistical software with relevant packages. Results The agreement rates between the scorer and AI-based automatic polysomnography scoring systems for each sleep stage (W, N1, N2, N3, R) were 89.96%, 77.93%, 91.13%, 84.04%, and 84.62%, respectively. The agreement rate between the scorer and AI for arousal was 82.5%, the agreement rate for respiratory events was 92.3%, and the agreement rate for periodic limb movements was 94.1%. Conclusion This study has demonstrated that SOMNUM is a valuable solution for automatic polysomnography interpretation. These results suggest that AI-based automated polysomnography scoring systems have the potential to advance the field of sleep medicine by improving diagnostic efficiency, reducing costs, and enhancing the consistency of sleep disorder diagnoses. Support (if any) This work was supported by the Seoul Business Agency (2023 Bio/Medical Technology Commercialization Supporting Project, BT230157). This work was supported by the Technology development Program (RS-2023-00321754) funded by the Ministry of SMEs and Startups (MSS, Korea).

Across-night oscillations of beta (20-28 Hz) and delta (0.3-3 Hz) electroencephalograms (EEGs) were examined with spectral analysis in 10 normal young adult subjects (Ss). In each S, power densities of beta were found to oscillate reciprocally with delta power density across both nonrapid eye movement (NREM) and rapid eye movement (REM) sleep. Linear correlation coefficients between log power density of delta vs. beta were significant (p less than 0.0001) for each S. An incidental observation was that beta power within REM was reliably lower in epochs with more eye movement activity. The reciprocal relationship between beta and delta holds implications for sleep physiology and supplements our earlier finding that sigma (12-15 Hz) oscillates reciprocally with delta within NREM sleep. These descriptions of the continuously varying EEG across sleep provide information not available when EEG measures are tabulated by discrete NREM periods and REM periods.

Our previous work showed a diminished cerebral blood flow (CBF) response to changes in Pa(CO(2)) in congestive heart failure patients with central sleep apnea compared with those without apnea. Since the regulation of CBF serves to minimize oscillations in H(+) and Pco(2) at the site of the central chemoreceptors, it may play an important role in maintaining breathing stability. We hypothesized that an attenuated cerebrovascular reactivity to changes in Pa(CO(2)) would narrow the difference between the eupneic Pa(CO(2)) and the apneic threshold Pa(CO(2)) (DeltaPa(CO(2))), known as the CO(2) reserve, thereby making the subjects more susceptible to apnea. Accordingly, in seven normal subjects, we used indomethacin (Indo; 100 mg by mouth) sufficient to reduce the CBF response to CO(2) by approximately 25% below control. The CO(2) reserve was estimated during non-rapid eye movement (NREM) sleep. The apnea threshold was determined, both with and without Indo, in NREM sleep, in a random order using a ventilator in pressure support mode to gradually reduce Pa(CO(2)) until apnea occurred. results: Indo significantly reduced the CO(2) reserve required to produce apnea from 6.3 +/- 0.5 to 4.4 +/- 0.7 mmHg (P = 0.01) and increased the slope of the ventilation decrease in response to hypocapnic inhibition below eupnea (control vs. Indo: 1.06 +/- 0.10 vs. 1.61 +/- 0.27 l x min(-1) x mmHg(-1), P < 0.05). We conclude that reductions in the normal cerebral vascular response to hypocapnia will increase the susceptibility to apneas and breathing instability during sleep.

Sleep and the Peripheral Vascular System

Don’t Let Your Patients Lose Sleep Over Insomnia

Although a large number of building roof reconstruction techniques from the remotely sensed data have been proposed, there is a serious lack in standard and automatic evaluation system. The manual evaluation systems are based on human judgment and so may be biased. Moreover, the application of such an evaluation system on a large data set requires a considerable amount of human involvement. In addition, the threshold-based evaluation systems that employ one or more thresholds while making correspondences between the detection and reference sets are also subjective since there is no unique way to choose the thresholds. This paper proposes an automatic and threshold-free evaluation system using the centre distances between the detected and reference entities. It neither requires any human involvement nor uses any thresholds. As a result, the proposed evaluation system can be exploited for a bias free evaluation of any roof reconstruction techniques, even on large data sets.