Tuomas Karhu

The traditional sleep staging involves manual scoring of electroencephalogram (EEG), electrooculogram (EOG), and electromyogram signals during polysomnography (PSG), which is laborious and susceptible to human errors. Previous studies have explored several automated sleep staging methods utilizing fewer EEG signals, for which deep-learning methods have shown promising results. Despite the availability of various signals and signal combinations in PSGs, the performance of different signal combinations for accurate sleep staging is not fully explored. We hypothesize that various EEG signal combinations will yield comparable performances, thus accurate automatic sleep staging could be achieved with a simplified measurement setup. We aim to identify the optimal EEG and EOG signal combinations for deep learning-based automatic sleep staging. Four EEG signals (F4-M1, C4-M1, O2-M1, and C3-M2) and one EOG signal (E1-M2) from 876 suspected obstructive sleep apnea subjects were studied. A total of 31 deep-learning models were trained utilizing different EEG and EOG signal combinations as input. The classification performance of automatic sleep staging against manual sleep staging was evaluated across five sleep stages. The differences in classification performance among various EEG signal combinations were negligible, with accuracies ranging from 81% to 85% (Cohen's kappa, κ = 0.73-0.78). Incorporating the EOG signal into single EEG configurations improved accuracies by 1-2 percentage points, while the improvements were smaller when combining EOG with multiple EEG signals. The comparable classification performances among various signal combinations suggest that automatic sleep staging can be achieved with a simplified EEG and EOG measurement setup without compromising performance.

Summary: Sleep stages exhibit differing patterns of cardiac, respiratory and nervous system activation. Rapid eye movement (REM) sleep is associated with deeper oxygen desaturation events in obstructive sleep apnea. However, no studies have looked to the effect of sleep staging on re‐saturation characteristics. Polysomnographic data from the Sleep Heart Health Study were utilised to derive oxygen saturation parameters from events having both desaturation and re‐saturation parts. Sleep stages were described as non‐REM (NREM) 1 or 2 (combined), 3 and REM. Sleep stage effects on desaturation and re‐saturation characteristics were investigated in a healthy subgroup ( n = 759) and participants with hypertension ( n = 2534), lung disease ( n = 715), heart failure ( n = 199) and myocardial infarction ( n = 713). A total of 3793 participants (48.3% female) were included for analysis (171,976 saturation events; median 20 per participant). Events during REM had the longest duration, deepest nadirs and greatest area for desaturations and re‐saturations. Sleep stage effected re‐saturation parameters more than desaturation parameters, with a relative difference from NREM 1/2 to REM in duration of 16.7% and 29.8%, in rate of 0.2% and 4.5% and in area of 36.1% and 48.0% for desaturation and re‐saturation, respectively, among healthy participants. Similarly, the effect of comorbidities was greater upon re‐saturations than desaturations, as participants with heart failure recorded a relative difference to healthy participants of 10.3% and 24.4% for desaturation and re‐saturation duration, respectively, 12.8% and 15.0% for rate and 6.4% and 16.4% for area. Sleep stages and comorbidities have significant effects upon nocturnal oxygen re‐saturation parameters. Cardiorespiratory comorbidities elicit greater degradations in oxygen re‐saturations than they do desaturations.

In obstructive sleep apnea (OSA), heart rate variability (HRV) decreases and performance in psychomotor vigilance task (PVT) worsens with more severe hypoxic load. Nevertheless, the association between HRV and PVT performance is poorly understood. Thus, we hypothesize that nocturnal short-term HRV is better related to daytime psychomotor vigilance compared with overnight HRV. To investigate this hypothesis, we retrospectively analyzed the electrocardiograms from polysomnographies of 546 consecutive patients with suspected OSA. We determined overnight HRV and short-term HRV in nonoverlapping 5-min segments and performed stepwise linear regression analyses to associate HRV with the median reaction time (RT) in the PVT. The short-term decrease in the median interval between two successive normal R peaks (NN interval), root mean square of successive NNs, and normalized high-frequency band power were all significant (p < 0.001) indicators of longer median RTs. However, the overnight HRV parameters did not indicate worsening median RT. Instead, increased hypoxic load and N3 duration were associated with longer median RT in men but not in women. The association of HRV and cardiorespiratory coupling with PVT performance was generally weak. Nocturnal short-term HRV evaluation reflected a state of vigilance better than the average overnight HRV. Thus, the overnight HRV analysis might not be optimal for patients with OSA. Utilizing the HRV analysis in a time-series manner and combined with the hypoxic load and sleep stages could bring new aspects to the health assessment of patients with OSA.

Summary: Obstructive sleep apnea is a heterogeneous sleep disorder with varying phenotypes. Several studies have already performed cluster analyses to discover various obstructive sleep apnea phenotypic clusters. However, the selection of the clustering method might affect the outputs. Consequently, it is unclear whether similar obstructive sleep apnea clusters can be reproduced using different clustering methods. In this study, we applied four well‐known clustering methods: Agglomerative Hierarchical Clustering; K‐means; Fuzzy C‐means; and Gaussian Mixture Model to a population of 865 suspected obstructive sleep apnea patients. By creating five clusters with each method, we examined the effect of clustering methods on forming obstructive sleep apnea clusters and the differences in their physiological characteristics. We utilized a visualization technique to indicate the cluster formations, Cohen's kappa statistics to find the similarity and agreement between clustering methods, and performance evaluation to compare the clustering performance. As a result, two out of five clusters were distinctly different with all four methods, while three other clusters exhibited overlapping features across all methods. In terms of agreement, Fuzzy C‐means and K‐means had the strongest (κ = 0.87), and Agglomerative hierarchical clustering and Gaussian Mixture Model had the weakest agreement (κ = 0.51) between each other. The K‐means showed the best clustering performance, followed by the Fuzzy C‐means in most evaluation criteria. Moreover, Fuzzy C‐means showed the greatest potential in handling overlapping clusters compared with other methods. In conclusion, we revealed a direct impact of clustering method selection on the formation and physiological characteristics of obstructive sleep apnea clusters. In addition, we highlighted the capability of soft clustering methods, particularly Fuzzy C‐means, in the application of obstructive sleep apnea phenotyping.

Objective: Hypoxic load is one of the main characteristics of obstructive sleep apnea (OSA) contributing to sympathetic overdrive and weakened cardiorespiratory coupling (CRC). Whether this association changes with increasing hypoxic load has remained obscure. Therefore, we aimed to study our hypothesis that increasing hypoxic load acutely decreases the CRC. Methods: We retrospectively analyzed the electrocardiography and nasal pressure signals in 5-min segment pairs ( n = 36 926) recorded during clinical polysomnographies of 603 patients with suspected OSA. The segment pairs were pooled into five groups based on the hypoxic load severity described with the the total integrated area under the blood oxygen saturation curve during desaturations. In these severity groups, we determined the frequency-domain heart rate variability (HRV) parameters, the HRV and respiratory high-frequency (HF, 0.15–0.4 Hz) peaks, and the difference between those peaks. We also computed the spectral HF coherence between HRV and respiration in the HF band. Results: The ratio of low-frequency (LF, 0.04–0.15 Hz) to HF power increased from 1.047 to 1.805 ( p < 0.001); the difference between the HRV and respiratory HF peaks increased from 0.001 Hz to 0.039 Hz ( p < 0.001); and the spectral coherence between HRV and respiration in the HF band decreased from 0.813 to 0.689 ( p < 0.001) as the hypoxic load increased. Conclusion and Significance: The vagal modulation decreases and CRC weakens significantly with increasing hypoxic load. Thus, the hypoxic load could be utilized more thoroughly in contemporary OSA diagnostics to better assess the severity of OSA-related cardiac stress.

Summary: Oxygen saturation (SpO₂)‐based parameters are more strongly linked to impaired daytime vigilance than the conventional diagnostic metrics in patients with obstructive sleep apnea (OSA). However, whether the association between SpO₂‐based parameters and impaired daytime vigilance is modulated by sex, remains unknown. Hence, we investigated the interplay between sex and detailed SpO₂‐based metrics and their association with impaired vigilance in patients with OSA. The study population consisted of 855 (473 males, 382 females) patients with suspected OSA who underwent overnight polysomnography and psychomotor vigilance task (PVT). The population was grouped by sex and divided into quartiles (Q1–Q4) based on median reaction times (RTs) in the PVT. In addition to conventional diagnostic metrics, desaturation severity (DesSev), fall severity (FallSev), and recovery severity (RecovSev) were compared between the sexes and between the best (Q1) and worst (Q4) performing quartiles by using cumulative distribution functions (CDFs). Additionally, sex‐specific covariate‐adjusted linear regression models were used to investigate the connection between the parameters and RTs. The CDFs showed significantly higher hypoxic load in Q4 in males compared to females. In addition, the DesSev (β = 8.05, p < 0.01), FallSev (β = 6.48, p = 0.02), RecovSev (β = 9.13, p < 0.01), and Oxygen Desaturation Index (β = 12.29, p < 0.01) were associated with increased RTs only in males. Conversely, the Arousal Index (β = 10.75–11.04, p < 0.01) was associated with impaired vigilance in females. The severity of intermittent hypoxaemia was strongly associated with longer RTs in males whereas the Arousal Index had the strongest association in females. Thus, the impact of hypoxic load on impaired vigilance seems to be stronger in males than females.

Obstructive sleep apnea (OSA) severity assessment is based on manually scored respiratory events and their arbitrary definitions. Thus, we present an alternative method to objectively evaluate OSA severity independently of the manual scorings and scoring rules. A retrospective envelope analysis was conducted on 847 suspected OSA patients. Four parameters were calculated from the difference between the nasal pressure signal’s upper and lower envelopes: average (AV), median (MD), standard deviation (SD), and coefficient of variation (CoV). We computed the parameters from the entirety of the recorded signals to perform binary classifications of patients using three different apnea–hypopnea index (AHI) thresholds (5-15-30). Additionally, the calculations were undertaken in 30-second epochs to estimate the ability of the parameters to detect manually scored respiratory events. Classification performances were assessed with areas under the curves (AUCs). As a result, the SD (AUCs ≥ 0.86) and CoV (AUCs ≥ 0.82) were the best classifiers for all AHI thresholds. Furthermore, non-OSA and severe OSA patients were separated well with SD (AUC = 0.97) and CoV (AUC = 0.95). Respiratory events within the epochs were identified moderately with MD (AUC = 0.76) and CoV (AUC = 0.82). In conclusion, envelope analysis is a promising alternative method by which to assess OSA severity without relying on manual scoring or the scoring rules of respiratory events.

Summary: Obstructive sleep apnea (OSA)‐related intermittent hypoxaemia is a potential risk factor for different OSA comorbidities, for example cardiovascular disease. However, conflicting results are found as to whether intermittent hypoxaemia is associated with impaired vigilance. Therefore, we aimed to investigate how desaturation characteristics differ between the non‐impaired vigilance and impaired vigilance patient groups formed based on psychomotor vigilance task (PVT) performance and compared with traditional OSA severity parameters. The study population comprised 863 patients with suspected OSA who underwent a PVT test before polysomnography. The conventional OSA parameters, for example, the apnea‐hypopnea index, oxygen desaturation index, and arousal index were computed. Furthermore, the median desaturation area, fall area, recovery area, and desaturation depth were computed with the pre‐event baseline reference and with reference to the 100% oxygen saturation level. Patients were grouped into best‐ and worst‐performing quartiles based on the number of lapses in PVT (Q1: PVT lapses 36). The association between parameters and impaired vigilance was evaluated by cumulative distribution functions (CDFs) and binomial logistic regression. Based on the CDFs, patients in Q4 had larger desaturation areas, recovery areas, and deeper desaturations when these were referenced to 100% saturation compared with Q1. The odds ratio (OR) of the median desaturation area (OR = 1.56), recovery area (OR = 1.71), and depth (OR = 1.65) were significantly elevated in Q4 in regression models. However, conventional OSA parameters were not significantly associated with impaired vigilance (ORs: 0.79–1.09). Considering desaturation parameters with a 100% SpO₂ reference in the diagnosis of OSA could provide additional information on the severity of OSA and related daytime vigilance impairment.