Automatic Respiratory Event Scoring in Obstructive Sleep Apnea Using a Long Short-Term Memory Neural Network.

The diagnosis of obstructive sleep apnea is based on daytime symptoms and the frequency of respiratory events during the night. The respiratory events are scored manually from polysomnographic recordings, which is time-consuming and expensive. Therefore, automatic scoring methods could considerably improve the efficiency of sleep apnea diagnostics and release the resources currently needed for manual scoring to other areas of sleep medicine. In this study, we trained a long short-term memory neural network for automatic scoring of respiratory events using input signals from peripheral blood oxygen saturation, thermistor-airflow, nasal pressure -airflow, and thorax respiratory effort. The signals were extracted from 887 in-lab polysomnography recordings. 787 patients with suspected sleep apnea were used to train the neural network and 100 patients were used as an independent test set. The epoch-wise agreement between manual and automatic neural network scoring was high (88.9%, κ = 0.728). In addition, the apnea-hypopnea index (AHI) calculated from the automated scoring was close to the manually determined AHI with a mean absolute error of 3.0 events/hour and an intraclass correlation coefficient of 0.985. The neural network approach for automatic scoring of respiratory events achieved high accuracy and good agreement with manual scoring. The presented neural network could be used for analysis of large research datasets that are unfeasible to score manually, and has potential for clinical use in the future In addition, since the neural network scores individual respiratory events, the automatic scoring can be easily reviewed manually if desired.

Neural network analysis of nocturnal SpO2 signal enables easy screening of sleep apnea in patients with acute cerebrovascular disease.

Current diagnostics of sleep apnea relies on the time-consuming manual analysis of complex sleep registrations, which is impractical for routine screening in hospitalized patients with a high probability for sleep apnea, e.g. those experiencing acute stroke or transient ischemic attacks (TIA). To overcome this shortcoming, we aimed to develop a convolutional neural network (CNN) capable of estimating the severity of sleep apnea in acute stroke and TIA patients based solely on the nocturnal oxygen saturation (SpO2) signal. The CNN was trained with SpO2 signals derived from 1379 home sleep apnea tests (HSAT) of suspected sleep apnea patients and tested with SpO2 signals of 77 acute ischemic stroke or TIA patients. The CNN's performance was tested by comparing the estimated respiratory event index (REI) and oxygen desaturation index (ODI) with manually obtained values. Median estimation errors for REI and ODI in patients with stroke or TIA were 1.45 events/hour and 0.61 events/hour, respectively. Furthermore, based on estimated REI and ODI, 77.9% and 88.3% of these patients were classified into the correct sleep apnea severity categories. The sensitivity and specificity to identify sleep apnea (REI > 5 events/hour) were 91.8% and 78.6%, respectively. Moderate-to-severe sleep apnea was detected (REI > 15 events/hour) with sensitivity of 92.3% and specificity of 96.1%. The CNN analysis of the SpO2 signal has great potential as a simple screening tool for sleep apnea. This novel automatic method accurately detects sleep apnea in acute cerebrovascular disease patients and facilitates their referral for a differential diagnostic HSAT or polysomnography evaluation.

Detailed Assessment of Sleep Architecture With Deep Learning and Shorter Epoch-to-Epoch Duration Reveals Sleep Fragmentation of Patients With Obstructive Sleep Apnea.

Traditional sleep staging with non-overlapping 30-second epochs overlooks multiple sleep-wake transitions. We aimed to overcome this by analyzing the sleep architecture in more detail with deep learning methods and hypothesized that the traditional sleep staging underestimates the sleep fragmentation of obstructive sleep apnea (OSA) patients. To test this hypothesis, we applied deep learning-based sleep staging to identify sleep stages with the traditional approach and by using overlapping 30-second epochs with 15-, 5-, 1-, or 0.5-second epoch-to-epoch duration. A dataset of 446 patients referred for polysomnography due to OSA suspicion was used to assess differences in the sleep architecture between OSA severity groups. The amount of wakefulness increased while REM and N3 decreased in severe OSA with shorter epoch-to-epoch duration. In other OSA severity groups, the amount of wake and N1 decreased while N3 increased. With the traditional 30-second epoch-to-epoch duration, only small differences in sleep continuity were observed between the OSA severity groups. With 1-second epoch-to-epoch duration, the hazard ratio illustrating the risk of fragmented sleep was 1.14 (p = 0.39) for mild OSA, 1.59 (p < 0.01) for moderate OSA, and 4.13 (p < 0.01) for severe OSA. With shorter epoch-to-epoch durations, total sleep time and sleep efficiency increased in the non-OSA group and decreased in severe OSA. In conclusion, more detailed sleep analysis emphasizes the highly fragmented sleep architecture in severe OSA patients which can be underestimated with traditional sleep staging. The results highlight the need for a more detailed analysis of sleep architecture when assessing sleep disorders.

Increased nocturnal arterial pulsation frequencies of obstructive sleep apnoea patients is associated with an increased number of lapses in a psychomotor vigilance task.

OBJECTIVES: Besides hypoxaemia severity, heart rate variability has been linked to cognitive decline in obstructive sleep apnoea (OSA) patients. Thus, our aim was to examine whether the frequency domain features of a nocturnal photoplethysmogram (PPG) can be linked to poor performance in the psychomotor vigilance task (PVT). METHODS: PPG signals from 567 suspected OSA patients, extracted from Type 1 diagnostic polysomnography, and corresponding results of PVT were retrospectively examined. The frequency content of complete PPGs was determined, and analyses were conducted separately for men (n=327) and women (n=240). Patients were grouped into PVT performance quartiles based on the number of lapses (reaction times ≥500 ms) and within-test variation in reaction times. The best-performing (Q1) and worst-performing (Q4) quartiles were compared due the lack of clinical thresholds in PVT. RESULTS: We found that the increase in arterial pulsation frequency (APF) in both men and women was associated with a higher number of lapses. Higher APF was also associated with higher within-test variation in men, but not in women. Median APF (ß=0.27, p=0.01), time spent under 90% saturation (ß=0.05, p<0.01), female sex (ß=1.29, p<0.01), older age (ß=0.03, p<0.01) and subjective sleepiness (ß=0.07, p<0.01) were significant predictors of belonging to Q4 based on lapses. Only female sex (ß=0.75, p<0.01) and depression (ß=0.91, p<0.02) were significant predictors of belonging to Q4 based on the within-test variation. CONCLUSIONS: In conclusion, increased APF in PPG provides a possible polysomnography indicator for deteriorated vigilance especially in male OSA patients. This finding highlights the connection between cardiorespiratory regulation, vigilance and OSA. However, our results indicate substantial sex-dependent differences that warrant further prospective studies.

Power spectral densities of nocturnal pulse oximetry signals differ in OSA patients with and without daytime sleepiness.

BACKGROUND: As nocturnal hypoxemia and heart rate variability are associated with excessive daytime sleepiness (EDS) related to OSA, we hypothesize that the power spectral densities (PSD) of nocturnal pulse oximetry signals could be utilized in the assessment of EDS. Thus, we aimed to investigate if PSDs contain features that are related to EDS and whether a convolutional neural network (CNN) could detect patients with EDS using self-learned PSD features. METHODS: A total of 915 OSA patients who had undergone polysomnography with multiple sleep latency test on the following day were investigated. PSDs for nocturnal blood oxygen saturation (SpO2), heart rate (HR), and photoplethysmogram (PPG), as well as power in the 15-35 mHz band in SpO2 (PSPO2) and HR (PHR), were computed. Differences in PSD features were investigated between EDS groups. Additionally, a CNN classifier was developed for identifying severe EDS patients based on spectral data. RESULTS: SpO2 power content increased significantly (p < 0.002) with increasing severity of EDS. Furthermore, a significant (p < 0.001) increase in HR-PSD was found in severe EDS (mean sleep latency < 5 min). Elevated odds of having severe EDS was found in PSPO2 (OR = 1.19-1.29) and PHR (OR = 1.81-1.83). Despite these significant spectral differences, the CNN classifier reached only moderate sensitivity (49.5%) alongside high specificity (80.4%) in identifying patients with severe EDS. CONCLUSIONS: We conclude that PSDs of nocturnal pulse oximetry signals contain features significantly associated with OSA-related EDS. However, CNN-based identification of patients with EDS is challenging via pulse oximetry.

Deep learning enables sleep staging from photoplethysmogram for patients with suspected sleep apnea.

STUDY OBJECTIVES: Accurate identification of sleep stages is essential in the diagnosis of sleep disorders (e.g. obstructive sleep apnea [OSA]) but relies on labor-intensive electroencephalogram (EEG)-based manual scoring. Furthermore, long-term assessment of sleep relies on actigraphy differentiating only between wake and sleep periods without identifying specific sleep stages and having low reliability in identifying wake periods after sleep onset. To address these issues, we aimed to develop an automatic method for identifying the sleep stages from the photoplethysmogram (PPG) signal obtained with a simple finger pulse oximeter. METHODS: PPG signals from the diagnostic polysomnographies of susptected OSA patients (n = 894) were utilized to develop a combined convolutional and recurrent neural network. The deep learning model was trained individually for three-stage (wake/NREM/REM), four-stage (wake/N1+N2/N3/REM), and five-stage (wake/N1/N2/N3/REM) classification of sleep. RESULTS: The three-stage model achieved an epoch-by-epoch accuracy of 80.1% with Cohen's κ of 0.65. The four- and five-stage models achieved 68.5% (κ = 0.54), and 64.1% (κ = 0.51) accuracies, respectively. With the five-stage model, the total sleep time was underestimated with a mean bias error (SD) of of 7.5 (55.2) minutes. CONCLUSION: The PPG-based deep learning model enabled accurate estimation of sleep time and differentiation between sleep stages with a moderate agreement to manual EEG-based scoring. As PPG is already included in ambulatory polygraphic recordings, applying the PPG-based sleep staging could improve their diagnostic value by enabling simple, low-cost, and reliable monitoring of sleep and help assess otherwise overlooked conditions such as REM-related OSA.

Estimating daytime sleepiness with previous night electroencephalography, electrooculography, and electromyography spectrograms in patients with suspected sleep apnea using a convolutional neural network.

A common symptom of obstructive sleep apnea (OSA) is excessive daytime sleepiness (EDS). The gold standard test for EDS is the multiple sleep latency test (MSLT). However, due to its high cost, MSLT is not routinely conducted for OSA patients and EDS is instead evaluated using sleep questionnaires. This is problematic however, since sleep questionnaires are subjective and correlate poorly with the MSLT. Therefore, new objective tools are needed for reliable evaluation of EDS. The aim of this study was to test our hypothesis that EDS can be estimated with neural network analysis of previous night polysomnographic signals. We trained a convolutional neural network (CNN) classifier using electroencephalography, electrooculography, and chin electromyography signals from 2,014 patients with suspected OSA. The CNN was trained to classify the patients into four sleepiness categories based on their mean sleep latency (MSL); severe (MSL < 5min), moderate (5 ≤ MSL < 10), mild (10 ≤ MSL < 15), and normal (MSL ≥ 15). The CNN classified patients to the four sleepiness categories with an overall accuracy of 60.6% and Cohen's kappa value of 0.464. In two-group classification scheme with sleepy (MSL < 10 min) and non-sleepy (MSL ≥ 10) patients, the CNN achieved an accuracy of 77.2%, with sensitivity of 76.5%, and specificity of 77.9%. Our results show that previous night's polysomnographic signals can be used for objective estimation of EDS with at least moderate accuracy. Since the diagnosis of OSA is currently confirmed by polysomnography, the classifier could be used simultaneously to get an objective estimate of the daytime sleepiness with minimal extra workload.

Severe desaturations increase psychomotor vigilance task-based median reaction time and number of lapses in obstructive sleep apnoea patients.

Current diagnostic parameters estimating obstructive sleep apnoea (OSA) severity have a poor connection to the psychomotor vigilance of OSA patients. Thus, we aimed to investigate how the severity of apnoeas, hypopnoeas and intermittent hypoxaemia is associated with impaired vigilance.We retrospectively examined type I polysomnography data and corresponding psychomotor vigilance tasks (PVTs) of 743 consecutive OSA patients (apnoea-hypopnoea index (AHI) ≥5 events·h-1). Conventional diagnostic parameters (e.g. AHI and oxygen desaturation index (ODI)) and novel parameters (e.g. desaturation severity and obstruction severity) incorporating duration of apnoeas and hypopnoeas as well as depth and duration of desaturations were assessed. Patients were grouped into quartiles based on PVT outcome variables. The odds of belonging to the worst-performing quartile were assessed. Analyses were performed for all PVT outcome variables using binomial logistic regression.A relative 10% increase in median depth of desaturations elevated the odds (ORrange 1.20-1.37, p<0.05) of prolonged mean and median reaction times as well as increased lapse count. Similarly, an increase in desaturation severity (ORrange 1.26-1.52, p<0.05) associated with prolonged median reaction time. Female sex (ORrange 2.21-6.02, p<0.01), Epworth Sleepiness Scale score (ORrange 1.05-1.07, p<0.01) and older age (ORrange 1.01-1.05, p<0.05) were significant risk factors in all analyses. In contrast, increases in conventional AHI, ODI and arousal index were not associated with deteriorated PVT performance.These results show that our novel parameters describing the severity of intermittent hypoxaemia are significantly associated with increased risk of impaired PVT performance, whereas conventional OSA severity and sleep fragmentation metrics are not. These results underline the importance of developing the assessment of OSA severity beyond the AHI.

Acute stroke and TIA patients have specific polygraphic features of obstructive sleep apnea.

PURPOSE: Obstructive sleep apnea (OSA) is associated with increased risk for stroke, which is known to further impair respiratory functions. However, it is unknown whether the type and severity of respiratory events are linked to stroke or transient ischemic attack (TIA). Thus, we investigate whether the characteristics of individual respiratory events differ between patients experiencing TIA or acute ischemic stroke and matched patients with clinically suspected sleep-disordered breathing. METHODS: Polygraphic data of 77 in-patients with acute ischemic stroke (n = 49) or TIA (n = 28) were compared to age, gender, and BMI-matched patients with suspected sleep-disordered breathing and no cerebrovascular disease. Along with conventional diagnostic parameters (e.g., apnea-hypopnea index), durations and severities of individual apneas, hypopneas and desaturations were compared between the groups separately for ischemic stroke and TIA patients. RESULTS: Stroke and TIA patients had significantly shorter apneas and hypopneas (p < 0.001) compared to matched reference patients. Furthermore, stroke patients had more central apnea events (p = 0.007) and a trend for higher apnea/hypopnea number ratios (p = 0.091). The prevalence of OSA (apnea-hypopnea index ≥ 5) was 90% in acute stroke patients and 79% in transient ischemic attack patients. CONCLUSION: Stroke patients had different characteristics of respiratory events, i.e., their polygraphic phenotype of OSA differs compared to matched reference patients. The observed differences in polygraphic features might indicate that stroke and TIA patients suffer from OSA phenotype recently associated with increased cardiovascular mortality. Therefore, optimal diagnostics and treatment require routine OSA screening in patients with acute cerebrovascular disease, even without previous suspicion of OSA.

Accurate Deep Learning-Based Sleep Staging in a Clinical Population With Suspected Obstructive Sleep Apnea.

The identification of sleep stages is essential in the diagnostics of sleep disorders, among which obstructive sleep apnea (OSA) is one of the most prevalent. However, manual scoring of sleep stages is time-consuming, subjective, and costly. To overcome this shortcoming, we aimed to develop an accurate deep learning approach for automatic classification of sleep stages and to study the effect of OSA severity on the classification accuracy. Overnight polysomnographic recordings from a public dataset of healthy individuals (Sleep-EDF, n = 153) and from a clinical dataset (n = 891) of patients with suspected OSA were used to develop a combined convolutional and long short-term memory neural network. On the public dataset, the model achieved sleep staging accuracy of 83.7% (κ = 0.77) with a single frontal EEG channel and 83.9% (κ = 0.78) when supplemented with EOG. For the clinical dataset, the model achieved accuracies of 82.9% (κ = 0.77) and 83.8% (κ = 0.78) with a single EEG channel and two channels (EEG+EOG), respectively. The sleep staging accuracy decreased with increasing OSA severity. The single-channel accuracy ranged from 84.5% (κ = 0.79) for individuals without OSA diagnosis to 76.5% (κ = 0.68) for patients with severe OSA. In conclusion, deep learning enables automatic sleep staging for suspected OSA patients with high accuracy and expectedly, the accuracy decreased with increasing OSA severity. Furthermore, the accuracies achieved in the public dataset were superior to previously published state-of-the-art methods. Adding an EOG channel did not significantly increase the accuracy. The automatic, single-channel-based sleep staging could enable easy, accurate, and cost-efficient integration of EEG recording into diagnostic ambulatory recordings.