Sleep staging from electrocardiography and respiration with deep learning

Improved sleep stage predictions by deep learning of photoplethysmogram and respiration patterns

Sleep staging is a crucial tool for diagnosing and monitoring sleep disorders, but the standard clinical approach using polysomnography (PSG) in a sleep lab is time-consuming, expensive, uncomfortable, and limited to a single night. Advancements in sensor technology have enabled home sleep monitoring, but existing devices still lack sufficient accuracy to inform clinical decisions. To address this challenge, we propose a deep learning architecture that combines a convolutional neural network and bidirectional long short-term memory to accurately classify sleep stages. By supplementing photoplethysmography (PPG) signals with respiratory sensor inputs, we demonstrated significant improvements in prediction accuracy and Cohen's kappa (k) for 2- (92.7 %; k = 0.768), 3- (80.2 %; k = 0.714), 4- (76.8 %, k = 0.550), and 5-stage (76.7 %, k = 0.616) sleep classification using raw data. This relatively translatable approach, with a less intensive AI model and leveraging only a few, inexpensive sensors, shows promise in accurately staging sleep. This has potential for diagnosing and managing sleep disorders in a more accessible and practical manner, possibly even at home.

Insights from the 2nd China intelligent sleep staging competition

STUDY OBJECTIVES

Artificial intelligence (AI) is quickly advancing in the field of sleep medicine, which bodes well for the potential of actual clinical use. In this study, an analysis of the 2nd China Intelligent Sleep Staging Competition was conducted to gain insights into the general level and constraints of AI-assisted sleep staging in China.

METHODS

The outcomes of 10 teams from the children's track and 13 teams from the adult track were investigated in this study. The analysis included overall performance, differences between five different sleep stages, variations across subjects, and performance during stage transitions.

RESULTS

The adult track's accuracy peaked at 80.46%, while the children's track's accuracy peaked at 88.96%. On average, accuracy rates stood at 71.43% for children and 68.40% for adults. All results were produced within a mere 5-min timeframe. The N1 stage was prone to misclassification as W, N2, and R stages. In the adult track, significant differences were apparent among subjects (p < 0.05), whereas in the children's track, such differences were not observed. Nonetheless, both tracks experienced a performance decline during stage transitions.

CONCLUSIONS

The computational speed of AI is remarkably fast, simultaneously holding the potential to surpass the accuracy of physicians. Improving the machine learning model's classification of the N1 stage and transitional periods between stages, along with bolstering its robustness to individual subject variations, is imperative for maximizing its ability in assisting clinical scoring.

Sleep staging algorithm based on smartwatch sensors for healthy and sleep apnea populations

OBJECTIVE

Sleep stages can provide valuable insights into an individual's sleep quality. By leveraging movement and heart rate data collected by modern smartwatches, it is possible to enable the sleep staging feature and enhance users' understanding about their sleep and health conditions.

METHOD

In this paper, we present and validate a recurrent neural network based model with 23 input features extracted from accelerometer and photoplethysmography sensors data for both healthy and sleep apnea populations. We designed a lightweight and fast solution to enable the prediction of sleep stages for each 30-s epoch. This solution was developed using a large dataset of 1522 night recordings collected from a highly heterogeneous population and different versions of Samsung smartwatch.

RESULTS

In the classification of four sleep stages (wake, light, deep, and rapid eye movements sleep), the proposed solution achieved 71.6 % of balanced accuracy and a Cohen's kappa of 0.56 in a test set with 586 recordings.

CONCLUSION

The results presented in this paper validate our proposal as a competitive wearable solution for sleep staging. Additionally, the use of a large and diverse data set contributes to the robustness of our solution, and corroborates the validation of algorithm's performance. Some additional analysis performed for healthy and sleep apnea population demonstrated that algorithm's performance has low correlation with demographic variables.

"Counting sheep PSG": EEGLAB-compatible open-source matlab software for signal processing, visualization, event marking and staging of polysomnographic data

BACKGROUND

Progress in advancing sleep research employing polysomnography (PSG) has been negatively impacted by the limited availability of widely available, open-source sleep-specific analysis tools.

NEW METHOD

Here, we introduce Counting Sheep PSG, an EEGLAB-compatible software for signal processing, visualization, event marking and manual sleep stage scoring of PSG data for MATLAB.

RESULTS

Key features include: (1) signal processing tools including bad channel interpolation, down-sampling, re-referencing, filtering, independent component analysis, artifact subspace reconstruction, and power spectral analysis, (2) customizable display of polysomnographic data and hypnogram, (3) event marking mode including manual sleep stage scoring, (4) automatic event detections including movement artifact, sleep spindles, slow waves and eye movements, and (5) export of main descriptive sleep architecture statistics, event statistics and publication-ready hypnogram.

COMPARISON WITH EXISTING METHODS

Counting Sheep PSG was built on the foundation created by sleepSMG (https://sleepsmg.sourceforge.net/). The scope and functionalities of the current software have made significant advancements in terms of EEGLAB integration/compatibility, preprocessing, artifact correction, event detection, functionality and ease of use. By comparison, commercial software can be costly and utilize proprietary data formats and algorithms, thereby restricting the ability to distribute and share data and analysis results.

CONCLUSIONS

The field of sleep research remains shackled by an industry that resists standardization, prevents interoperability, builds-in planned obsolescence, maintains proprietary black-box data formats and analysis approaches. This presents a major challenge for the field of sleep research. The need for free, open-source software that can read open-format data is essential for scientific advancement to be made in the field.

Diagnosing OSA and Insomnia at Home Based Only on an Actigraphy Total Sleep Time and RIP Belts an Algorithm "Nox Body Sleep™"

Purpose

The COVID-19 pandemic has influenced clinical sleep protocols with stricter hospital disinfection requirements. Facing these new rules, we tested if a new artificial intelligence (AI) algorithm: The Nox BodySleep™ (NBS) developed without airflow signals for the analysis of sleep might assess pertinently sleep in patients with Obstructive Sleep Apnea (OSA) and chronic insomnia (CI) as a control group, compared to polysomnography (PSG) manual scoring.

Patients-Methods

NBS is a recurrent neural network model that estimates Wake, NREM, and REM states, given features extracted from activity and respiratory inductance plethysmography (RIP) belt signals (Nox A1 PSG). Sleep states from 139 PSG studies (CI N = 72; OSA N = 67) were analyzed by NBS and compared to manually scored PSG using positive percentage agreement, negative percentage agreement, and overall agreement metrics. Similarly, we compared common sleep parameters and OSA severity using sleep states estimated by NBS for each recording and compared to manual scoring using Bland-Altman analysis and intra-class correlation coefficient.

Results

For 127,170 sleep epochs, an overall agreement of 83% was reached for Wake, NREM and REM states (92% for REM states in CI patients) between NBS and manually scored PSG. Overall agreement for estimating OSA severity was 100% for moderate-severe OSA and 91% for minimal OSA. The absolute errors of the apnea-hypopnea index (AHI) and total sleep time (TST) were significantly lower for the NBS compared to no scoring of sleep. The intra-class correlation was higher for AHI and significantly higher for TST using the NBS compared to no scoring of sleep.

Conclusion

NBS gives sleep states, parameters and AHI with a good positive and negative percentage agreement, compared with manually scored PSG.

Expert-level sleep staging using an electrocardiography-only feed-forward neural network

Reliable classification of sleep stages is crucial in sleep medicine and neuroscience research for providing valuable insights, diagnoses, and understanding of brain states. The current gold standard method for sleep stage classification is polysomnography (PSG). Unfortunately, PSG is an expensive and cumbersome process involving numerous electrodes, often conducted in an unfamiliar clinic and annotated by a professional. Although commercial devices like smartwatches track sleep, their performance is well below PSG. To address these disadvantages, we present a feed-forward neural network that achieves gold-standard levels of agreement using only a single lead of electrocardiography (ECG) data. Specifically, the median five-stage Cohen's kappa is 0.725 on a large, diverse dataset of 5 to 90-year-old subjects. Comparisons with a comprehensive meta-analysis of between-human inter-rater agreement confirm the non-inferior performance of our model. Finally, we developed a novel loss function to align the training objective with Cohen's kappa. Our method offers an inexpensive, automated, and convenient alternative for sleep stage classification-further enhanced by a real-time scoring option. Cardiosomnography, or a sleep study conducted with ECG only, could take expert-level sleep studies outside the confines of clinics and laboratories and into realistic settings. This advancement democratizes access to high-quality sleep studies, considerably enhancing the field of sleep medicine and neuroscience. It makes less-expensive, higher-quality studies accessible to a broader community, enabling improved sleep research and more personalized, accessible sleep-related healthcare interventions.

Modulate the impact of the drowsiness on the resting state functional connectivity

This research explores different methodologies to modulate the effects of drowsiness on functional connectivity (FC) during resting-state functional magnetic resonance imaging (RS-fMRI). The study utilized a cohort of students (MRi-Share) and classified individuals into drowsy, alert, and mixed/undetermined states based on observed respiratory oscillations. We analyzed the FC group difference between drowsy and alert individuals after five different processing methods: the reference method, two based on physiological and a global signal regression of the BOLD time series signal, and two based on Gaussian standardizations of the FC distribution. According to the reference method, drowsy individuals exhibit higher cortico-cortical FC than alert individuals. First, we demonstrated that each method reduced the differences between drowsy and alert states. The second result is that the global signal regression was quantitively the most effective, minimizing significant FC differences to only 3.3% of the total FCs. However, one should consider the risks of overcorrection often associated with this methodology. Therefore, choosing a less aggressive form of regression, such as the physiological method or Gaussian-based approaches, might be a more cautious approach. Third and last, using the Gaussian-based methods, cortico-subcortical and intra-default mode network (DMN) FCs were significantly greater in alert than drowsy subjects. These findings bear resemblance to the anticipated patterns during the onset of sleep, where the cortex isolates itself to assist in transitioning into deeper slow wave sleep phases, simultaneously disconnecting the DMN.

Research and application of deep learning-based sleep staging: Data, modeling, validation, and clinical practice

Over the past few decades, researchers have attempted to simplify and accelerate the process of sleep stage classification through various approaches; however, only a few such approaches have gained widespread acceptance. Artificial intelligence technology, particularly deep learning, is promising for earning the trust of the sleep medicine community in automated sleep-staging systems, thus facilitating its application in clinical practice and integration into daily life. We aimed to comprehensively review the latest methods that are applying deep learning for enhancing sleep staging efficiency and accuracy. Starting from the requisite "data" for constructing deep learning algorithms, we elucidated the current landscape of this domain and summarized the fundamental modeling process, encompassing signal selection, data pre-processing, model architecture, classification tasks, and performance metrics. Furthermore, we reviewed the applications of automated sleep staging in scenarios such as sleep-disorder screening, diagnostic procedures, and health monitoring and management. Finally, we conducted an in-depth analysis and discussion of the challenges and future in intelligent sleep staging, particularly focusing on large-scale sleep datasets, interdisciplinary collaborations, and human-computer interactions.

Developing a deep learning model for sleep stage prediction in obstructive sleep apnea cohort using 60 GHz frequency-modulated continuous-wave radar

Given the significant impact of sleep on overall health, radar technology offers a promising, non-invasive, and cost-effective avenue for the early detection of sleep disorders, even prior to relying on polysomnography (PSG)-based classification. In this study, we employed an attention-based bidirectional long short-term memory (Attention Bi-LSTM) model to accurately predict sleep stages using 60 GHz frequency-modulated continuous-wave (FMCW) radar. Our dataset comprised 78 participants from an ongoing obstructive sleep apnea (OSA) cohort, recruited between July 2021 and November 2022, who underwent overnight polysomnography alongside radar sensor monitoring. The dataset encompasses comprehensive polysomnography recordings, spanning both sleep and wakefulness states. The predictions achieved a Cohen's kappa coefficient of 0.746 and an overall accuracy of 85.2% in classifying wakefulness, rapid-eye-movement (REM) sleep, and non-REM (NREM) sleep (N1 + N2 + N3). The results demonstrated that the models incorporating both Radar 1 and Radar 2 data consistently outperformed those using only Radar 1 data, indicating the potential benefits of utilising multiple radars for sleep stage classification. Although the performance of the models tended to decline with increasing OSA severity, the addition of Radar 2 data notably improved the classification accuracy. These findings demonstrate the potential of radar technology as a valuable screening tool for sleep stage classification.

Open-source software for respiratory rate estimation using single-lead electrocardiograms

Respiratory rate (RR) is a critical vital sign used to assess pulmonary function. Currently, RR estimating instrumentation is specialized and bulky, therefore unsuitable for remote health monitoring. Previously, RR was estimated using proprietary software that extract surface electrocardiogram (ECG) waveform features obtained at several thoracic locations. However, developing a non-proprietary method that uses minimal ECG leads, generally available from mobile cardiac monitors is highly desirable. Here, we introduce an open-source and well-documented Python-based algorithm that estimates RR requiring only single-stream ECG signals. The algorithm was first developed using ECGs from awake, spontaneously breathing adult human subjects. The algorithm-estimated RRs exhibited close linear correlation to the subjects' true RR values demonstrating an R2 of 0.9092 and root mean square error of 2.2 bpm. The algorithm robustness was then tested using ECGs generated by the ischemic hearts of anesthetized, mechanically ventilated sheep. Although the ECG waveforms during ischemia exhibited severe morphologic changes, the algorithm-determined RRs exhibited high fidelity with a resolution of 1 bpm, an absolute error of 0.07 ± 0.07 bpm, and a relative error of 0.67 ± 0.64%. This optimized Python-based RR estimation technique will likely be widely adapted for remote lung function assessment in patients with cardiopulmonary disease.

Automatic sleep-stage classification of heart rate and actigraphy data using deep and transfer learning approaches

Manual sleep-stage scoring based on full-night polysomnography data recorded in a sleep lab has been the gold standard of clinical sleep medicine. This costly and time-consuming approach is unfit for long-term studies as well as assessment of sleep on a population level. With the vast amount of physiological data becoming available from wrist-worn devices, deep learning techniques provide an opportunity for fast and reliable automatic sleep-stage classification tasks. However, training a deep neural network requires large annotated sleep databases, which are not available for long-term epidemiological studies. In this paper, we introduce an end-to-end temporal convolutional neural network able to automatically score sleep stages from raw heartbeat RR interval (RRI) and wrist actigraphy data. Moreover, a transfer learning approach enables the training of the network on a large public database (Sleep Heart Health Study, SHHS) and its subsequent application to a much smaller database recorded by a wristband device. The transfer learning significantly shortens training time and improves sleep-scoring accuracy from 68.9% to 73.8% and inter-rater reliability (Cohen's kappa) from 0.51 to 0.59. We also found that for the SHHS database, automatic sleep-scoring accuracy using deep learning shows a logarithmic relationship with the training size. Although deep learning approaches for automatic sleep scoring are not yet comparable to the inter-rater reliability among sleep technicians, performance is expected to significantly improve in the near future when more large public databases become available. We anticipate those deep learning techniques, when combined with our transfer learning approach, will leverage automatic sleep scoring of physiological data from wearable devices and enable the investigation of sleep in large cohort studies.

Autonomic arousal detection and cardio-respiratory sleep staging improve the accuracy of home sleep apnea tests

Introduction: The apnea-hypopnea index (AHI), defined as the number of apneas and hypopneas per hour of sleep, is still used as an important index to assess sleep disordered breathing (SDB) severity, where hypopneas are confirmed by the presence of an oxygen desaturation or an arousal. Ambulatory polygraphy without neurological signals, often referred to as home sleep apnea testing (HSAT), can potentially underestimate the severity of sleep disordered breathing (SDB) as sleep and arousals are not assessed. We aim to improve the diagnostic accuracy of HSATs by extracting surrogate sleep and arousal information derived from autonomic nervous system activity with artificial intelligence. Methods: We used polysomnographic (PSG) recordings from 245 subjects (148 with simultaneously recorded HSATs) to develop and validate a new algorithm to detect autonomic arousals using artificial intelligence. A clinically validated auto-scoring algorithm (Somnolyzer) scored respiratory events, cortical arousals, and sleep stages in PSGs, and provided respiratory events and sleep stages from cardio-respiratory signals in HSATs. In a four-fold cross validation of the newly developed algorithm, we evaluated the accuracy of the estimated arousal index and HSAT-derived surrogates for the AHI. Results: The agreement between the autonomic and cortical arousal index was moderate to good with an intraclass correlation coefficient of 0.73. When using thresholds of 5, 15, and 30 to categorize SDB into none, mild, moderate, and severe, the addition of sleep and arousal information significantly improved the classification accuracy from 70.2% (Cohen's κ = 0.58) to 80.4% (κ = 0.72), with a significant reduction of patients where the severity category was underestimated from 18.8% to 7.3%. Discussion: Extracting sleep and arousal information from autonomic nervous system activity can improve the diagnostic accuracy of HSATs by significantly reducing the probability of underestimating SDB severity without compromising specificity.

High prevalence of sleep-disordered breathing in the intensive care unit - a cross-sectional study

PURPOSE

Sleep-disordered breathing may be induced by, exacerbate, or complicate recovery from critical illness. Disordered breathing during sleep, which itself is often fragmented, can go unrecognized in the intensive care unit (ICU). The objective of this study was to investigate the prevalence, severity, and risk factors of sleep-disordered breathing in ICU patients using a single respiratory belt and oxygen saturation signals.

METHODS

Patients in three ICUs at Massachusetts General Hospital wore a thoracic respiratory effort belt as part of a clinical trial for up to 7 days and nights. Using a previously developed machine learning algorithm, we processed respiratory and oximetry signals to measure the 3% apnea-hypopnea index (AHI) and estimate AH-specific hypoxic burden and periodic breathing. We trained models to predict AHI categories for 12-h segments from risk factors, including admission variables and bio-signals data, available at the start of these segments.

RESULTS

Of 129 patients, 68% had an AHI ≥ 5; 40% an AHI > 15, and 19% had an AHI > 30 while critically ill. Median [interquartile range] hypoxic burden was 2.8 [0.5, 9.8] at night and 4.2 [1.0, 13.7] %min/h during the day. Of patients with AHI ≥ 5, 26% had periodic breathing. Performance of predicting AHI-categories from risk factors was poor.

CONCLUSIONS

Sleep-disordered breathing and sleep apnea events while in the ICU are common and are associated with substantial burden of hypoxia and periodic breathing. Detection is feasible using limited bio-signals, such as respiratory effort and SpO2 signals, while risk factors were insufficient to predict AHI severity.

Clinical applications of artificial intelligence in sleep medicine: a sleep clinician's perspective

BACKGROUND

The past few years have seen a rapid emergence of artificial intelligence (AI)-enabled technology in the field of sleep medicine. AI refers to the capability of computer systems to perform tasks conventionally considered to require human intelligence, such as speech recognition, decision-making, and visual recognition of patterns and objects. The practice of sleep tracking and measuring physiological signals in sleep is widely practiced. Therefore, sleep monitoring in both the laboratory and ambulatory environments results in the accrual of massive amounts of data that uniquely positions the field of sleep medicine to gain from AI.

METHOD

The purpose of this article is to provide a concise overview of relevant terminology, definitions, and use cases of AI in sleep medicine. This was supplemented by a thorough review of relevant published literature.

RESULTS

Artificial intelligence has several applications in sleep medicine including sleep and respiratory event scoring in the sleep laboratory, diagnosing and managing sleep disorders, and population health. While still in its nascent stage, there are several challenges which preclude AI's generalizability and wide-reaching clinical applications. Overcoming these challenges will help integrate AI seamlessly within sleep medicine and augment clinical practice.

CONCLUSION

Artificial intelligence is a powerful tool in healthcare that may improve patient care, enhance diagnostic abilities, and augment the management of sleep disorders. However, there is a need to regulate and standardize existing machine learning algorithms prior to its inclusion in the sleep clinic.

Sleep staging in the ICU with heart rate variability and breathing signals. An exploratory cross-sectional study using deep neural networks

Introduction: To measure sleep in the intensive care unit (ICU), full polysomnography is impractical, while activity monitoring and subjective assessments are severely confounded. However, sleep is an intensely networked state, and reflected in numerous signals. Here, we explore the feasibility of estimating conventional sleep indices in the ICU with heart rate variability (HRV) and respiration signals using artificial intelligence methods Methods: We used deep learning models to stage sleep with HRV (through electrocardiogram) and respiratory effort (through a wearable belt) signals in critically ill adult patients admitted to surgical and medical ICUs, and in age and sex-matched sleep laboratory patients Results: We studied 102 adult patients in the ICU across multiple days and nights, and 220 patients in a clinical sleep laboratory. We found that sleep stages predicted by HRV- and breathing-based models showed agreement in 60% of the ICU data and in 81% of the sleep laboratory data. In the ICU, deep NREM (N2 + N3) proportion of total sleep duration was reduced (ICU 39%, sleep laboratory 57%, p < 0.01), REM proportion showed heavy-tailed distribution, and the number of wake transitions per hour of sleep (median 3.6) was comparable to sleep laboratory patients with sleep-disordered breathing (median 3.9). Sleep in the ICU was also fragmented, with 38% of sleep occurring during daytime hours. Finally, patients in the ICU showed faster and less variable breathing patterns compared to sleep laboratory patients Conclusion: The cardiovascular and respiratory networks encode sleep state information, which, together with artificial intelligence methods, can be utilized to measure sleep state in the ICU.

An automated heart rate-based algorithm for sleep stage classification: Validation using conventional polysomnography and an innovative wearable electrocardiogram device

Background

The rapid advancement in wearable solutions to monitor and score sleep staging has enabled monitoring outside of the conventional clinical settings. However, most of the devices and algorithms lack extensive and independent validation, a fundamental step to ensure robustness, stability, and replicability of the results beyond the training and testing phases. These systems are thought not to be feasible and reliable alternatives to the gold standard, polysomnography (PSG).

Materials and methods

This validation study highlights the accuracy and precision of the proposed heart rate (HR)-based deep-learning algorithm for sleep staging. The illustrated solution can perform classification at 2-levels (Wake; Sleep), 3-levels (Wake; NREM; REM) or 4- levels (Wake; Light; Deep; REM) in 30-s epochs. The algorithm was validated using an open-source dataset of PSG recordings (Physionet CinC dataset, n = 994 participants, 994 recordings) and a proprietary dataset of ECG recordings (Z3Pulse, n = 52 participants, 112 recordings) collected with a chest-worn, wireless sensor and simultaneous PSG collection using SOMNOtouch.

Results

We evaluated the performance of the models in both datasets in terms of Accuracy (A), Cohen's kappa (K), Sensitivity (SE), Specificity (SP), Positive Predictive Value (PPV), and Negative Predicted Value (NPV). In the CinC dataset, the highest value of accuracy was achieved by the 2-levels model (0.8797), while the 3-levels model obtained the best value of K (0.6025). The 4-levels model obtained the lowest SE (0.3812) and the highest SP (0.9744) for the classification of Deep sleep segments. AHI and biological sex did not affect scoring, while a significant decrease of performance by age was reported across the models. In the Z3Pulse dataset, the highest value of accuracy was achieved by the 2-levels model (0.8812), whereas the 3-levels model obtained the best value of K (0.611). For classification of the sleep states, the lowest SE (0.6163) and the highest SP (0.9606) were obtained for the classification of Deep sleep segment.

Conclusion

The results of the validation procedure demonstrated the feasibility of accurate HR-based sleep staging. The combination of the proposed sleep staging algorithm with an inexpensive HR device, provides a cost-effective and non-invasive solution deployable in the home environment and robust across age, sex, and AHI scores.

Automated sleep stage scoring employing a reasoning mechanism and evaluation of its explainability

Scoring sleep stages from biological signals is an essential but labor-intensive inspection for sleep diagnosis. The existing automated scoring methods have achieved high accuracy but are not widely applied in clinical practice. In our understanding, the existing methods have failed to establish the trust of sleep experts (e.g., physicians and clinical technologists) due to a lack of ability to explain the evidences/clues for scoring. In this study, we developed a deep-learning-based scoring model with a reasoning mechanism called class activation mapping (CAM) to solve this problem. This mechanism explicitly shows which portions of the signals support our model’s sleep stage decision, and we verified that these portions overlap with the “characteristic waves,” which are evidences/clues used in the manual scoring process. In exchange for the acquisition of explainability, employing CAM makes it difficult to follow some scoring rules. Although we concerned the negative effect of CAM on the scoring accuracy, we have found that the impact is limited. The evaluation experiment shows that the proposed model achieved a scoring accuracy of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$86.9\%$$\end{document}86.9%. It is superior to those of some existing methods and the inter-rater reliability among the sleep experts. These results suggest that Sleep-CAM achieved both explainability and required scoring accuracy for practical usage.

A systematic review of deep learning methods for modeling electrocardiograms during sleep

Sleep is one of the most important human physiological activities and plays an essential role in human health. Polysomnography (PSG) is the gold standard for measuring sleep quality and disorders, but it is time-consuming, labor-intensive, and prone to errors. Current research has confirmed the correlations between sleep and the respiratory/circulatory system. Electrocardiography (ECG) is convenient to perform, and ECG data are rich in breathing information. Therefore, sleep research based on ECG data has become popular. Currently, deep learning (DL) methods have achieved promising results on predictive health care tasks using ECG signals. Therefore, in this review, we systematically identify recent research studies and analyze them from the perspectives of data, model, and task. We discuss the shortcomings, summarize the findings, and highlight the potential opportunities. For sleep-related tasks, many ECG-based DL methods produce more accurate results than traditional approaches by combining multiple signal features and model structures. Methods that are more interpretable, scalable, and transferable will become ubiquitous in the daily practice of medicine and ambient-assisted-living applications. This paper is the first systematic review of ECG-based DL methods for sleep tasks.

The role of monoaminergic neurons in modulating respiration during sleep and the connection with SUDEP

Sudden unexpected death in epilepsy (SUDEP) is the leading cause of death among epilepsy patients, occurring even more frequently in cases with anti-epileptic drug resistance. Despite some advancements in characterizing SUDEP, the underlying mechanism remains incompletely understood. This review summarizes the latest advances in our understanding of the pathogenic mechanisms of SUDEP, in order to identify possible targets for the development of new strategies to prevent SUDEP. Based on our previous research along with the current literature, we focus on the role of sleep-disordered breathing (SDB) and its related neural mechanisms to consider the possible roles of monoaminergic neurons in the modulation of respiration during sleep and the occurrence of SUDEP. Overall, this review suggests that targeting the monoaminergic neurons is a promising approach to preventing SUDEP. The proposed roles of SDB and related monoaminergic neural mechanisms in SUDEP provide new insights for explaining the pathogenesis of SUDEP.

Jumping Knowledge Based Spatial-temporal Graph Convolutional Networks for Automatic Sleep Stage Classification

A novel jumping knowledge spatial-temporal graph convolutional network (JK-STGCN) is proposed in this paper to classify sleep stages. Based on this method, different types of multi-channel bio-signals, including electroencephalography (EEG), electromyogram (EMG), electrooculogram (EOG), and electrocardiogram (ECG) are utilized to classify sleep stages, after extracting features by a standard convolutional neural network (CNN) named FeatureNet. Intrinsic connections among different bio-signal channels from the identical epoch and neighboring epochs can be obtained through two adaptive adjacency matrices learning methods. A jumping knowledge spatial-temporal graph convolution module helps the JK-STGCN model to extract spatial features from the graph convolutions efficiently and temporal features are extracted from its common standard convolutions to learn the transition rules among sleep stages. Experimental results on the ISRUC-S3 dataset showed that the overall accuracy achieved 0.831 and the F1-score and Cohen kappa reached 0.814 and 0.782, respectively, which are the competitive classification performance with the state-of-the-art baselines. Further experiments on the ISRUC-S3 dataset are also conducted to evaluate the execution efficiency of the JK-STGCN model. The training time on 10 subjects is 2621s and the testing time on 50 subjects is 6.8s, which indicates its highest calculation speed compared with the existing high-performance graph convolutional networks and U-Net architecture algorithms. Experimental results on the ISRUC-S1 dataset also demonstrate its generality, whose accuracy, F1-score, and Cohen kappa achieve 0.820, 0.798, and 0.767 respectively.

Performance Evaluation of a Smart Bed Technology against Polysomnography

The Sleep Number smart bed uses embedded ballistocardiography, together with network connectivity, signal processing, and machine learning, to detect heart rate (HR), breathing rate (BR), and sleep vs. wake states. This study evaluated the performance of the smart bed relative to polysomnography (PSG) in estimating epoch-by-epoch HR, BR, sleep vs. wake, mean overnight HR and BR, and summary sleep variables. Forty-five participants (aged 22–64 years; 55% women) slept one night on the smart bed with standard PSG. Smart bed data were compared to PSG by Bland–Altman analysis and Pearson correlation for epoch-by-epoch HR and epoch-by-epoch BR. Agreement in sleep vs. wake classification was quantified using Cohen’s kappa, ROC analysis, sensitivity, specificity, accuracy, and precision. Epoch-by-epoch HR and BR were highly correlated with PSG (HR: r = 0.81, |bias| = 0.23 beats/min; BR: r = 0.71, |bias| = 0.08 breaths/min), as were estimations of mean overnight HR and BR (HR: r = 0.94, |bias| = 0.15 beats/min; BR: r = 0.96, |bias| = 0.09 breaths/min). Calculated agreement for sleep vs. wake detection included kappa (prevalence and bias-adjusted) = 0.74 ± 0.11, AUC = 0.86, sensitivity = 0.94 ± 0.05, specificity = 0.48 ± 0.18, accuracy = 0.86 ± 0.11, and precision = 0.90 ± 0.06. For all-night summary variables, agreement was moderate to strong. Overall, the findings suggest that the Sleep Number smart bed may provide reliable metrics to unobtrusively characterize human sleep under real life-conditions.

Evaluation of a Single-Channel EEG-Based Sleep Staging Algorithm

Sleep staging is the basis of sleep assessment and plays a crucial role in the early diagnosis and intervention of sleep disorders. Manual sleep staging by a specialist is time-consuming and is influenced by subjective factors. Moreover, some automatic sleep staging algorithms are complex and inaccurate. The paper proposes a single-channel EEG-based sleep staging method that provides reliable technical support for diagnosing sleep problems. In this study, 59 features were extracted from three aspects: time domain, frequency domain, and nonlinear indexes based on single-channel EEG data. Support vector machine, neural network, decision tree, and random forest classifier were used to classify sleep stages automatically. The results reveal that the random forest classifier has the best sleep staging performance among the four algorithms. The recognition rate of the Wake phase was the highest, at 92.13%, and that of the N1 phase was the lowest, at 73.46%, with an average accuracy of 83.61%. The embedded method was adopted for feature filtering. The results of sleep staging of the 11-dimensional features after filtering show that the random forest model achieved 83.51% staging accuracy under the condition of reduced feature dimensions, and the coincidence rate with the use of all features for sleep staging was 94.85%. Our study confirms the robustness of the random forest model in sleep staging, which also represents a high classification accuracy with appropriate classifier algorithms, even using single-channel EEG data. This study provides a new direction for the portability of clinical EEG monitoring.

Rhythmicity in heart rate and its surges usher a special period of sleep, a likely home for PGO waves

High amplitude electroencephalogram (EEG) events, like unitary K-complex (KC), are used to partition sleep into stages and hence define the hypnogram, a key instrument of sleep medicine. Throughout sleep the heart rate (HR) changes, often as a steady HR increase leading to a peak, what is known as a heart rate surge (HRS). The hypnogram is often unavailable when most needed, when sleep is disturbed and the graphoelements lose their identity. The hypnogram is also difficult to define during normal sleep, particularly at the start of sleep and the periods that precede and follow rapid eye movement (REM) sleep. Here, we use objective quantitative criteria that group together periods that cannot be assigned to a conventional sleep stage into what we call REM0 periods, with the presence of a HRS one of their defining properties. Extended REM0 periods are characterized by highly regular sequences of HRS that generate an infra-low oscillation around 0.05 Hz. During these regular sequence of HRS, and just before each HRS event, we find avalanches of high amplitude events for each one of the mass electrophysiological signals, i.e. related to eye movement, the motor system and the general neural activity. The most prominent features of long REM0 periods are sequences of three to five KCs which we label multiple K-complexes (KCm). Regarding HRS, a clear dissociation is demonstrated between the presence or absence of high gamma band spectral power (55-95 Hz) of the two types of KCm events: KCm events with strong high frequencies (KCmWSHF) cluster just before the peak of HRS, while KCm between HRS show no increase in high gamma band (KCmNOHF). Tomographic estimates of activity from magnetoencephalography (MEG) in pre-KC periods (single and multiple) showed common increases in the cholinergic Nucleus Basalis of Meynert in the alpha band. The direct contrast of KCmWSHF with KCmNOHF showed increases in all subjects in the high sigma band in the base of the pons and in three subjects in both the delta and high gamma bands in the medial Pontine Reticular Formation (mPRF), the putative Long Lead Initial pulse (LLIP) for Ponto-Geniculo-Occipital (PGO) waves.

Sleep Assessment in Critically Ill Adults: Established Methods and Emerging Strategies

Sleep is a biological mandate with an integral role in optimizing functions that maintain psychological and physical health. During critical illness, however, sleep may be disrupted at best and elusive at worst. Sleep improvement efforts and research endeavors evaluating interventions to improve sleep in critically ill adults are hampered by limited methods available to measure sleep in this setting. This narrative review summarizes available modalities for sleep assessment in the ICU, describes new ICU sleep assessment methods under development, and highlights features of the ideal ICU sleep measurement tool.

DATA SOURCES

The most relevant literature and author experiences were assessed for inclusion from PubMed and textbooks.

STUDY SELECTION

The authors selected studies for inclusion by consensus.

DATA EXTRACTION

The authors reviewed each study and selected appropriate data for inclusion by consensus.

DATA SYNTHESIS

Currently available tools to measure sleep in critically ill adults have important flaws. Subjective measurements are limited by recall bias, the inability of many patients to communicate, and poorly correlate with objective measures when completed by surrogates. Actigraphy does not consider the effects of sedating medications or myopathy leading to an over estimation of sleep time. Polysomnography, the gold standard for sleep assessment, is limited by interpretation issues and practical application concerns. Single and multiple channel electroencephalogram devices offer real-time physiologic data and are more practical to use than polysomnography but are limited by the scope of sleep-specific information they can measure and poorly characterize the circadian system.

CONCLUSIONS

A measurement tool that offers real-time sleep and circadian assessment and is practical for broad application in the ICU does not exist. Newer sleep assessment devices have shown promise in measuring physiologic data in real time; when used in combination with other assessment modalities, and analyzed by computational techniques, they may revolutionize sleep monitoring in the ICU.

A jerk-based algorithm ACCEL for the accurate classification of sleep–wake states from arm acceleration

Arm acceleration data have been used to measure sleep–wake rhythmicity. Although several methods have been developed for the accurate classification of sleep–wake episodes, a method with both high sensitivity and specificity has not been fully established. In this study, we developed an algorithm, named ACceleration-based Classification and Estimation of Long-term sleep–wake cycles (ACCEL) that classifies sleep and wake episodes using only raw accelerometer data, without relying on device-specific functions. The algorithm uses a derivative of triaxial acceleration (jerk), which can reduce individual differences in the variability of acceleration data. Applying a machine learning algorithm to the jerk data achieved sleep–wake classification with a high sensitivity (>90%) and specificity (>80%). A jerk-based analysis also succeeded in recording periodic activities consistent with pulse waves. Therefore, the ACCEL algorithm will be a useful method for large-scale sleep measurement using simple accelerometers in real-world settings.

•

An algorithm for sleep-wake classification based on arm acceleration is presented

•

The algorithm only uses a derivative of triaxial arm acceleration (jerk)

•

The algorithm can accurately detect temporal awake during sleep

End-to-End Sleep Staging Using Nocturnal Sounds from Microphone Chips for Mobile Devices

PURPOSE

Nocturnal sounds contain numerous information and are easily obtainable by a non-contact manner. Sleep staging using nocturnal sounds recorded from common mobile devices may allow daily at-home sleep tracking. The objective of this study is to introduce an end-to-end (sound-to-sleep stages) deep learning model for sound-based sleep staging designed to work with audio from microphone chips, which are essential in mobile devices such as modern smartphones.

PATIENTS AND METHODS

Two different audio datasets were used: audio data routinely recorded by a solitary microphone chip during polysomnography (PSG dataset, N=1154) and audio data recorded by a smartphone (smartphone dataset, N=327). The audio was converted into Mel spectrogram to detect latent temporal frequency patterns of breathing and body movement from ambient noise. The proposed neural network model learns to first extract features from each 30-second epoch and then analyze inter-epoch relationships of extracted features to finally classify the epochs into sleep stages.

RESULTS

Our model achieved 70% epoch-by-epoch agreement for 4-class (wake, light, deep, REM) sleep stage classification and robust performance across various signal-to-noise conditions. The model performance was not considerably affected by sleep apnea or periodic limb movement. External validation with smartphone dataset also showed 68% epoch-by-epoch agreement.

CONCLUSION

The proposed end-to-end deep learning model shows potential of low-quality sounds recorded from microphone chips to be utilized for sleep staging. Future study using nocturnal sounds recorded from mobile devices at home environment may further confirm the use of mobile device recording as an at-home sleep tracker.

Automated Scoring of Respiratory Events in Sleep with a Single Effort Belt and Deep Neural Networks

OBJECTIVE

Automatic detection and analysis of respiratory events in sleep using a single respiratory effort belt and deep learning.

METHODS

Using 9,656 polysomnography recordings from the Massachusetts General Hospital (MGH), we trained a neural network (WaveNet) to detect obstructive apnea, central apnea, hypopnea and respiratory-effort related arousals. Performance evaluation included event-based analysis and apnea-hypopnea index (AHI) stratification. The model was further evaluated on a public dataset, the Sleep-Heart-Health-Study-1, containing 8,455 polysomnographic recordings.

RESULTS

For binary apnea event detection in the MGH dataset, the neural network obtained a sensitivity of 68%, a specificity of 98%, a precision of 65%, a F1-score of 67%, and an area under the curve for the receiver operating characteristics curve and precision-recall curve of 0.93 and 0.71, respectively. AHI prediction resulted in a mean difference of 0.417.8 and a r² of 0.90. For the multiclass task, we obtained varying performances: 84% of all labeled central apneas were correctly classified, whereas this metric was 51% for obstructive apneas, 40% for respiratory effort related arousals and 23% for hypopneas.

CONCLUSION

Our fully automated method can detect respiratory events and assess the AHI accurately. Differentiation of event types is more difficult and may reflect in part the complexity of human respiratory output and some degree of arbitrariness in the criteria used during manual annotation.

SIGNIFICANCE

The current gold standard of diagnosing sleep-disordered breathing, using polysomnography and manual analysis, is time-consuming, expensive, and only applicable in dedicated clinical environments. Automated analysis using a single effort belt signal overcomes these limitations.

Enhancing Current Cardiorespiratory-based Approaches of Sleep Stage Classification by Temporal Feature Stacking

This paper presents a generic method to enhance performance and incorporate temporal information for cardiorespiratory-based sleep stage classification with a limited feature set and limited data. The classification algorithm relies on random forests and a feature set extracted from long-time home monitoring for sleep analysis. Employing temporal feature stacking, the system could be significantly improved in terms of Cohen's κ and accuracy. The detection performance could be improved for three classes of sleep stages (Wake, REM, Non-REM sleep), four classes (Wake, Non-REM-Light sleep, Non-REM Deep sleep, REM sleep), and five classes (Wake, N1, N2, N3/4, REM sleep) from a κ of 0.44 to 0.58, 0.33 to 0.51, and 0.28 to 0.44 respectively by stacking features before and after the epoch to be classified. Further analysis was done for the optimal length and combination method for this stacking approach. Overall, three methods and a variable duration between 30 s and 30 min have been analyzed. Overnight recordings of 36 healthy subjects from the Interdisciplinary Center for Sleep Medicine at Charité-Universitätsmedizin Berlin and Leave-One-Out-Cross-Validation on a patient-level have been used to validate the method.Clinical relevance- The method can be employed generically to feature sets for (small scale) datasets to improve classification performance for classification problems with temporal relations with random forest classifiers.

EEG Signal Multichannel Frequency-Domain Ratio Indices for Drowsiness Detection Based on Multicriteria Optimization

Drowsiness is a risk to human lives in many occupations and activities where full awareness is essential for the safe operation of systems and vehicles, such as driving a car or flying an airplane. Although it is one of the main causes of many road accidents, there is still no reliable definition of drowsiness or a system to reliably detect it. Many researchers have observed correlations between frequency-domain features of the EEG signal and drowsiness, such as an increase in the spectral power of the theta band or a decrease in the spectral power of the beta band. In addition, features calculated as ratio indices between these frequency-domain features show further improvements in detecting drowsiness compared to frequency-domain features alone. This work aims to develop novel multichannel ratio indices that take advantage of the diversity of frequency-domain features from different brain regions. In contrast to the state-of-the-art, we use an evolutionary metaheuristic algorithm to find the nearly optimal set of features and channels from which the indices are calculated. Our results show that drowsiness is best described by the powers in delta and alpha bands. Compared to seven existing single-channel ratio indices, our two novel six-channel indices show improvements in (1) statistically significant differences observed between wakefulness and drowsiness segments, (2) precision of drowsiness detection and classification accuracy of the XGBoost algorithm and (3) model performance by saving time and memory during classification. Our work suggests that a more precise definition of drowsiness is needed, and that accurate early detection of drowsiness should be based on multichannel frequency-domain features.

Sleep apnea and respiratory anomaly detection from a wearable band and oxygen saturation

OBJECTIVE

Sleep-related respiratory abnormalities are typically detected using polysomnography. There is a need in general medicine and critical care for a more convenient method to detect sleep apnea automatically from a simple, easy-to-wear device. The objective was to detect abnormal respiration and estimate the Apnea-Hypopnea Index (AHI) automatically with a wearable respiratory device with and without SpO₂ signals using a large (n = 412) dataset serving as ground truth.

DESIGN

Simultaneously recorded polysomnography (PSG) and wearable respiratory effort data were used to train and evaluate models in a cross-validation fashion. Time domain and complexity features were extracted, important features were identified, and a random forest model was employed to detect events and predict AHI. Four models were trained: one each using the respiratory features only, a feature from the SpO₂ (%)-signal only, and two additional models that use the respiratory features and the SpO₂ (%) feature, one allowing a time lag of 30 s between the two signals.

RESULTS

Event-based classification resulted in areas under the receiver operating characteristic curves of 0.94, 0.86, and 0.82, and areas under the precision-recall curves of 0.48, 0.32, and 0.51 for the models using respiration and SpO₂, respiration-only, and SpO₂-only, respectively. Correlation between expert-labelled and predicted AHI was 0.96, 0.78, and 0.93, respectively.

CONCLUSIONS

A wearable respiratory effort signal with or without SpO₂ signal predicted AHI accurately, and best performance was achieved with using both signals.

Transferring Structured Knowledge in Unsupervised Domain Adaptation of a Sleep Staging Network

Automatic sleep staging based on deep learning (DL) has been attracting attention for analyzing sleep quality and determining treatment effects. It is challenging to acquire long-term sleep data from numerous subjects and manually labeling them even though most DL-based models are trained using large-scale sleep data to provide state-of-the-art performance. One way to overcome this data shortage is to create a pre-trained network with an existing large-scale dataset (source domain) that is applicable to small cohorts of datasets (target domain); however, discrepancies in data distribution between the domains prevent successful refinement of this approach. In this paper, we propose an unsupervised domain adaptation method for sleep staging networks to reduce discrepancies by realigning the domains in the same space and producing domain-invariant features. Specifically, in addition to a classical domain discriminator, we introduce local dis-criminators-subject and stage-to maintain the intrinsic structure of sleep data to decrease local misalignments while using adversarial learning to play a minimax game between the feature extractor and discriminators. Moreover, we present several optimization schemes during training because the conventional adversarial learning is not effective to our training scheme. We evaluate the performance of the proposed method by examining the staging performances of a baseline network compared with direct transfer (DT) learning in various conditions. The experimental results demonstrate that the proposed domain adaptation significantly improves the performance though it needs no labeled sleep data in target domain.

Automatic sleep staging by cardiorespiratory signals: a systematic review

BACKGROUND

Because of problems with the recording and analysis of the EEG signal, automatic sleep staging using cardiorespiratory signals has been employed as an alternative. This study reports on certain critical points which hold considerable promise for the improvement of the results of the automatic sleep staging using cardiorespiratory signals.

METHODS

A systematic review.

RESULTS

The review and analysis of the literature in this area revealed four outstanding points: (1) the feature extraction epoch length, denoting that the standard 30-s segments of cardiorespiratory signals do not carry enough information for automatic sleep staging and that a 4.5-min length segment centering on each 30-s segment is proper for staging, (2) the time delay between the EEG signal extracted from the central nervous system activity and the cardiorespiratory signals extracted from the autonomic nervous system activity should be considered in the automatic sleep staging using cardiorespiratory signals, (3) the information in the morphology of ECG signals can contribute to the improvement of sleep staging, and (4) applying convolutional neural network (CNN) and long short-term memory network (LSTM) deep structures simultaneously to a large PSG recording database can lead to more reliable automatic sleep staging results.

CONCLUSIONS

Considering the above-mentioned points simultaneously can improve automatic sleep staging by cardiorespiratory signals. It is hoped that by considering the points, staging sleep automatically using cardiorespiratory signals, which does not have problems with the recording and analysis of EEG signals, yields results acceptably close to the results of automatic sleep staging by EEG signals.

Estimating sleep stages using cardiorespiratory signals: validation of a novel algorithm across a wide range of sleep-disordered breathing severity

STUDY OBJECTIVES

We have developed the CardioRespiratory Sleep Staging (CReSS) algorithm for estimating sleep stages using heart rate variability and respiration, allowing for estimation of sleep staging during home sleep apnea tests. Our objective was to undertake an epoch-by-epoch validation of algorithm performance against the gold standard of manual polysomnography sleep staging.

METHODS

Using 296 polysomnographs, we created a limited montage of airflow and heart rate and deployed CReSS to identify each 30-second epoch as wake, light sleep (N1 + N2), deep sleep (N3), or rapid eye movement (REM) sleep. We calculated Cohen's kappa and the percentage of accurately identified epochs. We repeated our analyses after stratification by sleep-disordered breathing (SDB) severity, and after adding thoracic respiratory effort as a backup signal for periods of invalid airflow.

RESULTS

CReSS discriminated wake/light sleep/deep sleep/REM sleep with 78% accuracy; the kappa value was 0.643 (95% confidence interval, 0.641-0.645). Discrimination of wake/sleep demonstrated a kappa value of 0.711 and accuracy of 89%, non-REM sleep/REM sleep demonstrated a kappa of 0.790 and accuracy of 94%, and light sleep/deep sleep demonstrated a kappa of 0.469 and accuracy of 87%. Kappa values did not vary by more than 0.07 across subgroups of no SDB, mild SDB, moderate SDB, and severe SDB. Accuracy increased to 80%, with a kappa value of 0.680 (95% confidence interval, 0.678-0.682), when CReSS additionally utilized the thoracic respiratory effort signal.

CONCLUSIONS

We observed substantial agreement between CReSS and the gold-standard comparator of manual sleep staging of polysomnographic signals, which was consistent across the full range of SDB severity. Future research should focus on the extent to which CReSS reduces the discrepancy between the apnea-hypopnea index and the respiratory event index, and the ability of CReSS to identify REM sleep-related obstructive sleep apnea.

Automatic Sleep Stage Classification of Children with Sleep-Disordered Breathing Using the Modularized Network

PURPOSE

To develop an automatic sleep stage analysis model for children and evaluate the effect of the model on the diagnosis of sleep-disordered breathing (SDB).

PATIENTS AND METHODS

Three hundred and forty-four SDB patients aged between 2 to 18 years who completed polysomnography (PSG) to assess the severity of the disease were enrolled in this study. We developed deep neural networks to stage sleep from electroencephalography (EEG), electrooculography (EOG) and electromyogram (EMG). The model performance was estimated by accuracy, precision, recall, F1-score, and Cohen's Kappa coefficient (ĸ). And we compared the difference in calculation of sleep parameters among the technicians, the model ensemble, and the single-channel EEG model.

RESULTS

The numbers of raw data divided into training, validation, and testing were 240, 36, and 68, respectively. The best performance appeared in the model ensemble of which the accuracy was 83.36% (ĸ=0.7817) in 5-stages, and the accuracy was 96.76% (ĸ=0.8236) in 2-stages. The single-channel EEG model showed the classification satisfyingly as well. There was no significant difference in TST, SE, SOL, time in W, time in N1+N2, time in N3, and OAHI between technician and the model (P>0.05). On the datasets from sleep-EDF-13 and sleep-EDF-18, the average classification accuracies achieved were 92.76% and 91.94% in 5-stages by using the proposed method, respectively.

CONCLUSION

This research established the model for pediatric automatic sleep stage classification with satisfying reliability and generalizability. In addition, it could be applied for calculating quantitative sleep parameters and evaluating the severity of SDB.

Machine and Deep Learning in Molecular and Genetic Aspects of Sleep Research

Epidemiological sleep research strives to identify the interactions and causal mechanisms by which sleep affects human health, and to design intervention strategies for improving sleep throughout the lifespan. These goals can be advanced by further focusing on the environmental and genetic etiology of sleep disorders, and by development of risk stratification algorithms, to identify people who are at risk or are affected by, sleep disorders. These studies rely on comprehensive sleep-related data which often contains complex multi-dimensional physiological and molecular measurements across multiple timepoints. Thus, sleep research is well-suited for the application of computational approaches that can handle high-dimensional data. Here, we survey recent advances in machine and deep learning together with the availability of large human cohort studies with sleep data that can jointly drive the next breakthroughs in the sleep-research field. We describe sleep-related data types and datasets, and present some of the tasks in the field that can be targets for algorithmic approaches, as well as the challenges and opportunities in pursuing them.

Confidence-Based Framework Using Deep Learning for Automated Sleep Stage Scoring

STUDY OBJECTIVES

Automated sleep stage scoring is not yet vigorously used in practice because of the black-box nature and the risk of wrong predictions. The objective of this study was to introduce a confidence-based framework to detect the possibly wrong predictions that would inform clinicians about which epochs would require a manual review and investigate the potential to improve accuracy for automated sleep stage scoring.

METHODS

We used 702 polysomnography studies from a local clinical dataset (SNUBH dataset) and 2804 from an open dataset (SHHS dataset) for experiments. We adapted the state-of-the-art TinySleepNet architecture to train the classifier and modified the ConfidNet architecture to train an auxiliary confidence model. For the confidence model, we developed a novel method, Dropout Correct Rate (DCR), and the performance of it was compared with other existing methods.

RESULTS

Confidence estimates (0.754) reflected accuracy (0.758) well in general. The best performance for differentiating correct and wrong predictions was shown when using the DCR method (AUROC: 0.812) compared to the existing approaches which largely failed to detect wrong predictions. By reviewing only 20% of epochs that received the lowest confidence values, the overall accuracy of sleep stage scoring was improved from 76% to 87%. For patients with reduced accuracy (ie, individuals with obesity or severe sleep apnea), the possible improvement range after applying confidence estimation was even greater.

CONCLUSION

To the best of our knowledge, this is the first study applying confidence estimation on automated sleep stage scoring. Reliable confidence estimates by the DCR method help screen out most of the wrong predictions, which would increase the reliability and interpretability of automated sleep stage scoring.

Direct application of an ECG-based sleep staging algorithm on reflective photoplethysmography data decreases performance

OBJECTIVE

The maturation of neural network-based techniques in combination with the availability of large sleep datasets has increased the interest in alternative methods of sleep monitoring. For unobtrusive sleep staging, the most promising algorithms are based on heart rate variability computed from inter-beat intervals (IBIs) derived from ECG-data. The practical application of these algorithms is even more promising when alternative ways of obtaining IBIs, such as wrist-worn photoplethysmography (PPG) can be used. However, studies validating sleep staging algorithms directly on PPG-based data are limited.

RESULTS

We applied an automatic sleep staging algorithm trained and validated on ECG-data directly on inter-beat intervals derived from a wrist-worn PPG sensor, in 389 polysomnographic recordings of patients with a variety of sleep disorders. While the algorithm reached moderate agreement with gold standard polysomnography, the performance was significantly lower when applied on PPG- versus ECG-derived heart rate variability data (kappa 0.56 versus 0.60, p < 0.001; accuracy 73.0% versus 75.9% p < 0.001). These results show that direct application of an algorithm on a different source of data may negatively affect performance. Algorithms need to be validated using each data source and re-training should be considered whenever possible.

IGRNet: A Deep Learning Model for Non-Invasive, Real-Time Diagnosis of Prediabetes through Electrocardiograms

The clinical symptoms of prediabetes are mild and easy to overlook, but prediabetes may develop into diabetes if early intervention is not performed. In this study, a deep learning model—referred to as IGRNet—is developed to effectively detect and diagnose prediabetes in a non-invasive, real-time manner using a 12-lead electrocardiogram (ECG) lasting 5 s. After searching for an appropriate activation function, we compared two mainstream deep neural networks (AlexNet and GoogLeNet) and three traditional machine learning algorithms to verify the superiority of our method. The diagnostic accuracy of IGRNet is 0.781, and the area under the receiver operating characteristic curve (AUC) is 0.777 after testing on the independent test set including mixed group. Furthermore, the accuracy and AUC are 0.856 and 0.825, respectively, in the normal-weight-range test set. The experimental results indicate that IGRNet diagnoses prediabetes with high accuracy using ECGs, outperforming existing other machine learning methods; this suggests its potential for application in clinical practice as a non-invasive, prediabetes diagnosis technology.