Multichannel Sleep Stage Classification and Transfer Learning using Convolutional Neural Networks

Optimizing wearable single-channel electroencephalography sleep staging in a heterogeneous sleep-disordered population using transfer learning

STUDY OBJECTIVES

Although various wearable electroencephalography devices have been developed, performance evaluation of the devices and their associated automated sleep stage classification models is mostly limited to healthy participants. A major barrier for applying automated wearable electroencephalography sleep staging in clinical populations is the need for large-scale data for model training. We therefore investigated transfer learning as a strategy to overcome limited data availability and optimize automated single-channel electroencephalography sleep staging in people with sleep disorders.

METHODS

We acquired 52 single-channel frontopolar headband electroencephalography recordings from a heterogeneous sleep-disordered population with concurrent polysomnography (PSG). We compared 3 model training strategies: "pretraining" (ie, training on a larger dataset of 901 conventional PSGs), "training-from-scratch" (ie, training on wearable headband recordings), and "fine-tuning" (ie, training on conventional PSGs, followed by training on headband recordings). Performance was evaluated on all headband recordings using 10-fold cross-validation.

RESULTS

Highest performance for 5-stage classification was achieved with fine-tuning (κ = .778), significantly higher than with pretraining (κ = .769) and with training-from-scratch (κ = .733). No significant differences or systematic biases were observed with clinically relevant sleep parameters derived from PSG. All sleep disorder categories showed comparable performance.

CONCLUSIONS

This study emphasizes the importance of leveraging larger available datasets through deep transfer learning to optimize performance with limited data availability. Findings indicate strong similarity in data characteristics between conventional PSG and headband recordings. Altogether, results suggest the combination of the headband, classification model, and training methodology can be viable for sleep monitoring in the heterogeneous clinical population.

CITATION

van der Aar JF, van Gilst MM, van den Ende DA, et al. Optimizing wearable single-channel electroencephalography sleep staging in a heterogeneous sleep-disordered population using transfer learning. J Clin Sleep Med. 2025;21(2):315-323.

Harnessing Few-Shot Learning for EEG signal classification: a survey of state-of-the-art techniques and future directions

This paper presents a systematic literature review, providing a comprehensive taxonomy of Data Augmentation (DA), Transfer Learning (TL), and Self-Supervised Learning (SSL) techniques within the context of Few-Shot Learning (FSL) for EEG signal classification. EEG signals have shown significant potential in various paradigms, including Motor Imagery, Emotion Recognition, Visual Evoked Potentials, Steady-State Visually Evoked Potentials, Rapid Serial Visual Presentation, Event-Related Potentials, and Mental Workload. However, challenges such as limited labeled data, noise, and inter/intra-subject variability have impeded the effectiveness of traditional machine learning (ML) and deep learning (DL) models. This review methodically explores how FSL approaches, incorporating DA, TL, and SSL, can address these challenges and enhance classification performance in specific EEG paradigms. It also delves into the open research challenges related to these techniques in EEG signal classification. Specifically, the review examines the identification of DA strategies tailored to various EEG paradigms, the creation of TL architectures for efficient knowledge transfer, and the formulation of SSL methods for unsupervised representation learning from EEG data. Addressing these challenges is crucial for enhancing the efficacy and robustness of FSL-based EEG signal classification. By presenting a structured taxonomy of FSL techniques and discussing the associated research challenges, this systematic review offers valuable insights for future investigations in EEG signal classification. The findings aim to guide and inspire researchers, promoting advancements in applying FSL methodologies for improved EEG signal analysis and classification in real-world settings.

Machine learning-empowered sleep staging classification using multi-modality signals

The goal is to enhance an automated sleep staging system's performance by leveraging the diverse signals captured through multi-modal polysomnography recordings. Three modalities of PSG signals, namely electroencephalogram (EEG), electrooculogram (EOG), and electromyogram (EMG), were considered to obtain the optimal fusions of the PSG signals, where 63 features were extracted. These include frequency-based, time-based, statistical-based, entropy-based, and non-linear-based features. We adopted the ReliefF (ReF) feature selection algorithms to find the suitable parts for each signal and superposition of PSG signals. Twelve top features were selected while correlated with the extracted feature sets' sleep stages. The selected features were fed into the AdaBoost with Random Forest (ADB + RF) classifier to validate the chosen segments and classify the sleep stages. This study's experiments were investigated by obtaining two testing schemes: epoch-wise testing and subject-wise testing. The suggested research was conducted using three publicly available datasets: ISRUC-Sleep subgroup1 (ISRUC-SG1), sleep-EDF(S-EDF), Physio bank CAP sleep database (PB-CAPSDB), and S-EDF-78 respectively. This work demonstrated that the proposed fusion strategy overestimates the common individual usage of PSG signals.

Simplifying Multimodal With Single EOG Modality for Automatic Sleep Staging

Polysomnography (PSG) recordings have been widely used for sleep staging in clinics, containing multiple modality signals (i.e., EEG and EOG). Recently, many studies have combined EEG and EOG modalities for sleep staging, since they are the most and the second most powerful modality for sleep staging among PSG recordings, respectively. However, EEG is complex to collect and sensitive to environment noise or other body activities, imbedding its use in clinical practice. Comparatively, EOG is much more easily to be obtained. In order to make full use of the powerful ability of EEG and the easy collection of EOG, we propose a novel framework to simplify multimodal sleep staging with a single EOG modality. It still performs well with only EOG modality in the absence of the EEG. Specifically, we first model the correlation between EEG and EOG, and then based on the correlation we generate multimodal features with time and frequency guided generators by adopting the idea of generative adversarial learning. We collected a real-world sleep dataset containing 67 recordings and used other four public datasets for evaluation. Compared with other existing sleep staging methods, our framework performs the best when solely using the EOG modality. Moreover, under our framework, EOG provides a comparable performance to EEG.

Sleep-Deep-Net learns sleep wake scoring from the end-user and completes each record in their style

Sleep-wake scoring is a time-consuming, tedious but essential component of clinical and pre-clinical sleep research. Sleep scoring is even more laborious and challenging in rodents due to the smaller EEG amplitude differences between states and the rapid state transitions which necessitate scoring in shorter epochs. Although many automated rodent sleep scoring methods exist, they do not perform as well when scoring new data sets, especially those which involve changes in the EEG/EMG profile. Thus, manual scoring by expert scorers remains the gold-standard. Here we take a different approach to this problem by using a neural network to accelerate the scoring of expert scorers. Sleep-Deep-Net (SDN) creates a bespoke deep convolution neural network model for individual electroencephalographic or local-field-potential records via transfer learning of GoogleNet, by learning from a small subset of manual scores of each EEG/LFP record as provided by the end-user. SDN then automates scoring of the remainder of the EEG/LFP record. A novel REM scoring correction procedure further enhanced accuracy. SDN reliably scores EEG and LFP data and retains sleep-wake architecture in wild-type mice, in sleep induced by the hypnotic zolpidem, in a mouse model of Alzheimer's disease and in a genetic knock-down study, when compared to manual scoring. SDN reduced manual scoring time to 1/12. Since SDN uses transfer learning on each independent recording, it is not biased by previously scored existing data sets. Thus, we find SDN performs well when used on signals altered by a drug, disease model or genetic modification.

Few-shot EEG sleep staging based on transductive prototype optimization network

Electroencephalography (EEG) is a commonly used technology for monitoring brain activities and diagnosing sleep disorders. Clinically, doctors need to manually stage sleep based on EEG signals, which is a time-consuming and laborious task. In this study, we propose a few-shot EEG sleep staging termed transductive prototype optimization network (TPON) method, which aims to improve the performance of EEG sleep staging. Compared with traditional deep learning methods, TPON uses a meta-learning algorithm, which generalizes the classifier to new classes that are not visible in the training set, and only have a few examples for each new class. We learn the prototypes of existing objects through meta-training, and capture the sleep features of new objects through the "learn to learn" method of meta-learning. The prototype distribution of the class is optimized and captured by using support set and unlabeled high confidence samples to increase the authenticity of the prototype. Compared with traditional prototype networks, TPON can effectively solve too few samples in few-shot learning and improve the matching degree of prototypes in prototype network. The experimental results on the public SleepEDF-2013 dataset show that the proposed algorithm outperform than most advanced algorithms in the overall performance. In addition, we experimentally demonstrate the feasibility of cross-channel recognition, which indicates that there are many similar sleep EEG features between different channels. In future research, we can further explore the common features among different channels and investigate the combination of universal features in sleep EEG. Overall, our method achieves high accuracy in sleep stage classification, demonstrating the effectiveness of this approach and its potential applications in other medical fields.

A convolutional neural network-based decision support system for neonatal quiet sleep detection

Sleep plays an important role in neonatal brain and physical development, making its detection and characterization important for assessing early-stage development. In this study, we propose an automatic and computationally efficient algorithm to detect neonatal quiet sleep (QS) using a convolutional neural network (CNN). Our study used 38-hours of electroencephalography (EEG) recordings, collected from 19 neonates at Fudan Children's Hospital in Shanghai, China (Approval No. (2020) 22). To train and test the CNN, we extracted 12 prominent time and frequency domain features from 9 bipolar EEG channels. The CNN architecture comprised two convolutional layers with pooling and rectified linear unit (ReLU) activation. Additionally, a smoothing filter was applied to hold the sleep stage for 3 minutes. Through performance testing, our proposed method achieved impressive results, with 94.07% accuracy, 89.70% sensitivity, 94.40% specificity, 79.82% F1-score and a 0.74 kappa coefficient when compared to human expert annotations. A notable advantage of our approach is its computational efficiency, with the entire training and testing process requiring only 7.97 seconds. The proposed algorithm has been validated using leave one subject out (LOSO) validation, which demonstrates its consistent performance across a diverse range of neonates. Our findings highlight the potential of our algorithm for real-time neonatal sleep stage classification, offering a fast and cost-effective solution. This research opens avenues for further investigations in early-stage development monitoring and the assessment of neonatal health.

Comparison of deep transfer learning algorithms and transferability measures for wearable sleep staging

Background

Obtaining medical data using wearable sensors is a potential replacement for in-hospital monitoring, but the lack of data for such sensors poses a challenge for development. One solution is using in-hospital recordings to boost performance via transfer learning. While there are many possible transfer learning algorithms, few have been tested in the domain of EEG-based sleep staging. Furthermore, there are few ways for determining which transfer learning method will work best besides exhaustive testing. Measures of transferability do exist, but are typically used for selection of pre-trained models rather than algorithms and few have been tested on medical signals. We tested several supervised transfer learning algorithms on a sleep staging task using a single channel of EEG (AF7-Fpz) captured from an in-home commercial system.

Results

Two neural networks—one bespoke and another state-of-art open-source architecture—were pre-trained on one of six source datasets comprising 11,561 subjects undergoing clinical polysomnograms (PSGs), then re-trained on a target dataset of 75 full-night recordings from 24 subjects. Several transferability measures were then tested to determine which is most effective for assessing performance on unseen target data. Performance on the target dataset was improved using transfer learning, with re-training the head layers being the most effective in the majority of cases (up to 63.9% of cases). Transferability measures generally provided significant correlations with accuracy (up to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r = -0.53$$\end{document}r=-0.53).

Conclusion

Re-training the head layers provided the largest performance boost. Transferability measures are useful indicators of transfer learning effectiveness.

XSleepNet: Multi-View Sequential Model for Automatic Sleep Staging

Automating sleep staging is vital to scale up sleep assessment and diagnosis to serve millions experiencing sleep deprivation and disorders and enable longitudinal sleep monitoring in home environments. Learning from raw polysomnography signals and their derived time-frequency image representations has been prevalent. However, learning from multi-view inputs (e.g., both the raw signals and the time-frequency images) for sleep staging is difficult and not well understood. This work proposes a sequence-to-sequence sleep staging model, XSleepNet,¹ that is capable of learning a joint representation from both raw signals and time-frequency images. Since different views may generalize or overfit at different rates, the proposed network is trained such that the learning pace on each view is adapted based on their generalization/overfitting behavior. In simple terms, the learning on a particular view is speeded up when it is generalizing well and slowed down when it is overfitting. View-specific generalization/overfitting measures are computed on-the-fly during the training course and used to derive weights to blend the gradients from different views. As a result, the network is able to retain the representation power of different views in the joint features which represent the underlying distribution better than those learned by each individual view alone. Furthermore, the XSleepNet architecture is principally designed to gain robustness to the amount of training data and to increase the complementarity between the input views. Experimental results on five databases of different sizes show that XSleepNet consistently outperforms the single-view baselines and the multi-view baseline with a simple fusion strategy. Finally, XSleepNet also outperforms prior sleep staging methods and improves previous state-of-the-art results on the experimental databases.

A multi-modal assessment of sleep stages using adaptive Fourier decomposition and machine learning

Healthy sleep is essential for the rejuvenation of the body and helps in maintaining good health. Many people suffer from sleep disorders that are characterized by abnormal sleep patterns. Automated assessment of such disorders using biomedical signals has been an active subject of research. Electroencephalogram (EEG) is a popular diagnostic used in this regard. We consider a widely-used publicly available database and process the signals using the Fourier decomposition method (FDM) to obtain narrowband signal components. Statistical features extracted from these components are passed on to machine learning classifiers to identify different stages of sleep. A novel feature measuring the non-stationarity of the signal is also used to capture salient information. It is shown that classification results can be improved by using multi-channel EEG instead of single-channel EEG data. Simultaneous utilization of multiple modalities, such as Electromyogram (EMG), Electrooculogram (EOG) along with EEG data leads to further enhancement in the obtained results. The proposed method can be efficiently implemented in real-time using fast Fourier transform (FFT), and it provides better classification results than the other algorithms existing in the literature. It can assist in the development of low-cost sensor-based setups for continuous patient monitoring and feedback.

Automatic sleep staging of EEG signals: recent development, challenges, and future directions

Modern deep learning holds a great potential to transform clinical practice on human sleep. Teaching a machine to carry out routine tasks would be a tremendous reduction in workload for clinicians. Sleep staging, a fundamental step in sleep practice, is a suitable task for this and will be the focus in this article. Recently, automatic sleep staging systems have been trained to mimic manual scoring, leading to similar performance to human sleep experts, at least on scoring of healthy subjects. Despite tremendous progress, we have not seen automatic sleep scoring adopted widely in clinical environments. This review aims to give a shared view of the authors on the most recent state-of-the-art development in automatic sleep staging, the challenges that still need to be addressed, and the future directions for automatic sleep scoring to achieve clinical value.

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.

A Novel Sleep Staging Network Based on Data Adaptation and Multimodal Fusion

Sleep staging is one of the important methods to diagnosis and treatment of sleep diseases. However, it is laborious and time-consuming, therefore, computer assisted sleep staging is necessary. Most of the existing sleep staging researches using hand-engineered features rely on prior knowledges of sleep analysis, and usually single channel electroencephalogram (EEG) is used for sleep staging task. Prior knowledge is not always available, and single channel EEG signal cannot fully represent the patient’s sleeping physiological states. To tackle the above two problems, we propose an automatic sleep staging network model based on data adaptation and multimodal feature fusion using EEG and electrooculogram (EOG) signals. 3D-CNN is used to extract the time-frequency features of EEG at different time scales, and LSTM is used to learn the frequency evolution of EOG. The nonlinear relationship between the High-layer features of EEG and EOG is fitted by deep probabilistic network. Experiments on SLEEP-EDF and a private dataset show that the proposed model achieves state-of-the-art performance. Moreover, the prediction result is in accordance with that from the expert diagnosis.

Attention based convolutional network for automatic sleep stage classification

Sleep staging is an important basis for diagnosing sleep-related problems. In this paper, an attention based convolutional network for automatic sleep staging is proposed. The network takes time-frequency image as input and predict sleep stage for each 30-s epoch as output. For each CNN feature maps, our model generate attention maps along two separate dimensions, time and filter, and then multiplied to form the final attention map. Residual-like fusion structure is used to append the attention map to the input feature map for adaptive feature refinement. In addition, to get the global feature representation with less information loss, the generalized mean pooling is introduced. To prove the efficacy of the proposed method, we have compared with two baseline method on sleep-EDF data set with different setting of the framework and input channel type, the experimental results show that the paper model has achieved significant improvements in terms of overall accuracy, Cohen's kappa, MF1, sensitivity and specificity. The performance of the proposed network is compared with that of the state-of-the-art algorithms with an overall accuracy of 83.4%, a macro F1-score of 77.3%, κ = 0.77, sensitivity = 77.1% and specificity = 95.4%, respectively. The experimental results demonstrate the superiority of the proposed network.

RobustSleepNet: Transfer Learning for Automated Sleep Staging at Scale

Sleep disorder diagnosis relies on the analysis of polysomnography (PSG) records. As a preliminary step of this examination, sleep stages are systematically determined. In practice, sleep stage classification relies on the visual inspection of 30-second epochs of polysomnography signals. Numerous automatic approaches have been developed to replace this tedious and expensive task. Although these methods demonstrated better performance than human sleep experts on specific datasets, they remain largely unused in sleep clinics. The main reason is that each sleep clinic uses a specific PSG montage that most automatic approaches cannot handle out-of-the-box. Moreover, even when the PSG montage is compatible, publications have shown that automatic approaches perform poorly on unseen data with different demographics. To address these issues, we introduce RobustSleepNet, a deep learning model for automatic sleep stage classification able to handle arbitrary PSG montages. We trained and evaluated this model in a leave-one-out-dataset fashion on a large corpus of 8 heterogeneous sleep staging datasets to make it robust to demographic changes. When evaluated on an unseen dataset, RobustSleepNet reaches 97% of the F1 of a model explicitly trained on this dataset. Hence, RobustSleepNet unlocks the possibility to perform high-quality out-of-the-box automatic sleep staging with any clinical setup. We further show that finetuning RobustSleepNet, using a part of the unseen dataset, increases the F1 by 2% when compared to a model trained specifically for this dataset. Therefore, finetuning might be used to reach a state-of-the-art level of performance on a specific population.

Interpretable deep learning for the remote characterisation of ambulation in multiple sclerosis using smartphones

The emergence of digital technologies such as smartphones in healthcare applications have demonstrated the possibility of developing rich, continuous, and objective measures of multiple sclerosis (MS) disability that can be administered remotely and out-of-clinic. Deep Convolutional Neural Networks (DCNN) may capture a richer representation of healthy and MS-related ambulatory characteristics from the raw smartphone-based inertial sensor data than standard feature-based methodologies. To overcome the typical limitations associated with remotely generated health data, such as low subject numbers, sparsity, and heterogeneous data, a transfer learning (TL) model from similar large open-source datasets was proposed. Our TL framework leveraged the ambulatory information learned on human activity recognition (HAR) tasks collected from wearable smartphone sensor data. It was demonstrated that fine-tuning TL DCNN HAR models towards MS disease recognition tasks outperformed previous Support Vector Machine (SVM) feature-based methods, as well as DCNN models trained end-to-end, by upwards of 8-15%. A lack of transparency of "black-box" deep networks remains one of the largest stumbling blocks to the wider acceptance of deep learning for clinical applications. Ensuing work therefore aimed to visualise DCNN decisions attributed by relevance heatmaps using Layer-Wise Relevance Propagation (LRP). Through the LRP framework, the patterns captured from smartphone-based inertial sensor data that were reflective of those who are healthy versus people with MS (PwMS) could begin to be established and understood. Interpretations suggested that cadence-based measures, gait speed, and ambulation-related signal perturbations were distinct characteristics that distinguished MS disability from healthy participants. Robust and interpretable outcomes, generated from high-frequency out-of-clinic assessments, could greatly augment the current in-clinic assessment picture for PwMS, to inform better disease management techniques, and enable the development of better therapeutic interventions.

MetaSleepLearner: A Pilot Study on Fast Adaptation of Bio-Signals-Based Sleep Stage Classifier to New Individual Subject Using Meta-Learning

Identifying bio-signals based-sleep stages requires time-consuming and tedious labor of skilled clinicians. Deep learning approaches have been introduced in order to challenge the automatic sleep stage classification conundrum. However, the difficulties can be posed in replacing the clinicians with the automatic system due to the differences in many aspects found in individual bio-signals, causing the inconsistency in the performance of the model on every incoming individual. Thus, we aim to explore the feasibility of using a novel approach, capable of assisting the clinicians and lessening the workload. We propose the transfer learning framework, entitled MetaSleepLearner, based on Model Agnostic Meta-Learning (MAML), in order to transfer the acquired sleep staging knowledge from a large dataset to new individual subjects (source code is available at https://github.com/IoBT-VISTEC/MetaSleepLearner). The framework was demonstrated to require the labelling of only a few sleep epochs by the clinicians and allow the remainder to be handled by the system. Layer-wise Relevance Propagation (LRP) was also applied to understand the learning course of our approach. In all acquired datasets, in comparison to the conventional approach, MetaSleepLearner achieved a range of 5.4% to 17.7% improvement with statistical difference in the mean of both approaches. The illustration of the model interpretation after the adaptation to each subject also confirmed that the performance was directed towards reasonable learning. MetaSleepLearner outperformed the conventional approaches as a result from the fine-tuning using the recordings of both healthy subjects and patients. This is the first work that investigated a non-conventional pre-training method, MAML, resulting in a possibility for human-machine collaboration in sleep stage classification and easing the burden of the clinicians in labelling the sleep stages through only several epochs rather than an entire recording.

A Survey on Deep Learning-Based Short/Zero-Calibration Approaches for EEG-Based Brain-Computer Interfaces

Brain-computer interfaces (BCIs) utilizing machine learning techniques are an emerging technology that enables a communication pathway between a user and an external system, such as a computer. Owing to its practicality, electroencephalography (EEG) is one of the most widely used measurements for BCI. However, EEG has complex patterns and EEG-based BCIs mostly involve a cost/time-consuming calibration phase; thus, acquiring sufficient EEG data is rarely possible. Recently, deep learning (DL) has had a theoretical/practical impact on BCI research because of its use in learning representations of complex patterns inherent in EEG. Moreover, algorithmic advances in DL facilitate short/zero-calibration in BCI, thereby suppressing the data acquisition phase. Those advancements include data augmentation (DA), increasing the number of training samples without acquiring additional data, and transfer learning (TL), taking advantage of representative knowledge obtained from one dataset to address the so-called data insufficiency problem in other datasets. In this study, we review DL-based short/zero-calibration methods for BCI. Further, we elaborate methodological/algorithmic trends, highlight intriguing approaches in the literature, and discuss directions for further research. In particular, we search for generative model-based and geometric manipulation-based DA methods. Additionally, we categorize TL techniques in DL-based BCIs into explicit and implicit methods. Our systematization reveals advances in the DA and TL methods. Among the studies reviewed herein, ~45% of DA studies used generative model-based techniques, whereas ~45% of TL studies used explicit knowledge transferring strategy. Moreover, based on our literature review, we recommend an appropriate DA strategy for DL-based BCIs and discuss trends of TLs used in DL-based BCIs.

Towards More Accurate Automatic Sleep Staging via Deep Transfer Learning

BACKGROUND

Despite recent significant progress in the development of automatic sleep staging methods, building a good model still remains a big challenge for sleep studies with a small cohort due to the data-variability and data-inefficiency issues. This work presents a deep transfer learning approach to overcome these issues and enable transferring knowledge from a large dataset to a small cohort for automatic sleep staging.

METHODS

We start from a generic end-to-end deep learning framework for sequence-to-sequence sleep staging and derive two networks as the means for transfer learning. The networks are first trained in the source domain (i.e. the large database). The pretrained networks are then finetuned in the target domain (i.e. the small cohort) to complete knowledge transfer. We employ the Montreal Archive of Sleep Studies (MASS) database consisting of 200 subjects as the source domain and study deep transfer learning on three different target domains: the Sleep Cassette subset and the Sleep Telemetry subset of the Sleep-EDF Expanded database, and the Surrey-cEEGrid database. The target domains are purposely adopted to cover different degrees of data mismatch to the source domains.

RESULTS

Our experimental results show significant performance improvement on automatic sleep staging on the target domains achieved with the proposed deep transfer learning approach.

CONCLUSIONS

These results suggest the efficacy of the proposed approach in addressing the above-mentioned data-variability and data-inefficiency issues.

SIGNIFICANCE

As a consequence, it would enable one to improve the quality of automatic sleep staging models when the amount of data is relatively small.¹¹The source code and the pretrained models are published at https://github.com/pquochuy/sleep_transfer_learning.

Automatic sleep stage classification with deep residual networks in a mixed-cohort setting

STUDY OBJECTIVES

Sleep stage scoring is performed manually by sleep experts and is prone to subjective interpretation of scoring rules with low intra- and interscorer reliability. Many automatic systems rely on few small-scale databases for developing models, and generalizability to new datasets is thus unknown. We investigated a novel deep neural network to assess the generalizability of several large-scale cohorts.

METHODS

A deep neural network model was developed using 15,684 polysomnography studies from five different cohorts. We applied four different scenarios: (1) impact of varying timescales in the model; (2) performance of a single cohort on other cohorts of smaller, greater, or equal size relative to the performance of other cohorts on a single cohort; (3) varying the fraction of mixed-cohort training data compared with using single-origin data; and (4) comparing models trained on combinations of data from 2, 3, and 4 cohorts.

RESULTS

Overall classification accuracy improved with increasing fractions of training data (0.25%: 0.782 ± 0.097, 95% CI [0.777-0.787]; 100%: 0.869 ± 0.064, 95% CI [0.864-0.872]), and with increasing number of data sources (2: 0.788 ± 0.102, 95% CI [0.787-0.790]; 3: 0.808 ± 0.092, 95% CI [0.807-0.810]; 4: 0.821 ± 0.085, 95% CI [0.819-0.823]). Different cohorts show varying levels of generalization to other cohorts.

CONCLUSIONS

Automatic sleep stage scoring systems based on deep learning algorithms should consider as much data as possible from as many sources available to ensure proper generalization. Public datasets for benchmarking should be made available for future research.

U-Sleep: resilient high-frequency sleep staging

Sleep disorders affect a large portion of the global population and are strong predictors of morbidity and all-cause mortality. Sleep staging segments a period of sleep into a sequence of phases providing the basis for most clinical decisions in sleep medicine. Manual sleep staging is difficult and time-consuming as experts must evaluate hours of polysomnography (PSG) recordings with electroencephalography (EEG) and electrooculography (EOG) data for each patient. Here, we present U-Sleep, a publicly available, ready-to-use deep-learning-based system for automated sleep staging ( sleep.ai.ku.dk ). U-Sleep is a fully convolutional neural network, which was trained and evaluated on PSG recordings from 15,660 participants of 16 clinical studies. It provides accurate segmentations across a wide range of patient cohorts and PSG protocols not considered when building the system. U-Sleep works for arbitrary combinations of typical EEG and EOG channels, and its special deep learning architecture can label sleep stages at shorter intervals than the typical 30 s periods used during training. We show that these labels can provide additional diagnostic information and lead to new ways of analyzing sleep. U-Sleep performs on par with state-of-the-art automatic sleep staging systems on multiple clinical datasets, even if the other systems were built specifically for the particular data. A comparison with consensus-scores from a previously unseen clinic shows that U-Sleep performs as accurately as the best of the human experts. U-Sleep can support the sleep staging workflow of medical experts, which decreases healthcare costs, and can provide highly accurate segmentations when human expertize is lacking.

Deep Neural Networks and Transfer Learning on a Multivariate Physiological Signal Dataset

While Deep Neural Networks (DNNs) and Transfer Learning (TL) have greatly contributed to several medical and clinical disciplines, the application to multivariate physiological datasets is still limited. Current examples mainly focus on one physiological signal and can only utilise applications that are customised for that specific measure, thus it limits the possibility of transferring the trained DNN to other domains. In this study, we composed a dataset (n=813) of six different types of physiological signals (Electrocardiogram, Electrodermal activity, Electromyogram, Photoplethysmogram, Respiration and Acceleration). Signals were collected from 232 subjects using four different acquisition devices. We used a DNN to classify the type of physiological signal and to demonstrate how the TL approach allows the exploitation of the efficiency of DNNs in other domains. After the DNN was trained to optimally classify the type of signal, the features that were automatically extracted by the DNN were used to classify the type of device used for the acquisition using a Support Vector Machine. The dataset, the code and the trained parameters of the DNN are made publicly available to encourage the adoption of DNN and TL in applications with multivariate physiological signals.

Multi-Branch Convolutional Neural Network for Automatic Sleep Stage Classification with Embedded Stage Refinement and Residual Attention Channel Fusion

Automatic sleep stage classification of multi-channel sleep signals can help clinicians efficiently evaluate an individual's sleep quality and assist in diagnosing a possible sleep disorder. To obtain accurate sleep classification results, the processing flow of results from signal preprocessing and machine-learning-based classification is typically employed. These classification results are refined based on sleep transition rules. Neural networks-i.e., machine learning algorithms-are powerful at solving classification problems. Some methods apply them to the first two processes above; however, the refinement process continues to be based on traditional methods. In this study, the sleep stage refinement process was incorporated into the neural network model to form real end-to-end processing. In addition, for multi-channel signals, the multi-branch convolutional neural network was combined with a proposed residual attention method. This approach further improved the model classification accuracy. The proposed method was evaluated on the Sleep-EDF Expanded Database (Sleep-EDFx) and University College Dublin Sleep Apnea Database (UCDDB). It achieved respective accuracy rates of 85.7% and 79.4%. The results also showed that sleep stage refinement based on a neural network is more effective than the traditional refinement method. Moreover, the proposed residual attention method was determined to have a more robust channel-information fusion ability than the respective average and concatenation methods.

Development of Automated Sleep Stage Classification System Using Multivariate Projection-Based Fixed Boundary Empirical Wavelet Transform and Entropy Features Extracted from Multichannel EEG Signals

The categorization of sleep stages helps to diagnose different sleep-related ailments. In this paper, an entropy-based information–theoretic approach is introduced for the automated categorization of sleep stages using multi-channel electroencephalogram (EEG) signals. This approach comprises of three stages. First, the decomposition of multi-channel EEG signals into sub-band signals or modes is performed using a novel multivariate projection-based fixed boundary empirical wavelet transform (MPFBEWT) filter bank. Second, entropy features such as bubble and dispersion entropies are computed from the modes of multi-channel EEG signals. Third, a hybrid learning classifier based on class-specific residuals using sparse representation and distances from nearest neighbors is used to categorize sleep stages automatically using entropy-based features computed from MPFBEWT domain modes of multi-channel EEG signals. The proposed approach is evaluated using the multi-channel EEG signals obtained from the cyclic alternating pattern (CAP) sleep database. Our results reveal that the proposed sleep staging approach has obtained accuracies of 91.77%, 88.14%, 80.13%, and 73.88% for the automated categorization of wake vs. sleep, wake vs. rapid eye movement (REM) vs. Non-REM, wake vs. light sleep vs. deep sleep vs. REM sleep, and wake vs. S1-sleep vs. S2-sleep vs. S3-sleep vs. REM sleep schemes, respectively. The developed method has obtained the highest overall accuracy compared to the state-of-art approaches and is ready to be tested with more subjects before clinical application.

Convolution-and Attention-Based Neural Network for Automated Sleep Stage Classification

Analyzing polysomnography (PSG) is an effective method for evaluating sleep health; however, the sleep stage scoring required for PSG analysis is a time-consuming effort for an experienced medical expert. When scoring sleep epochs, experts pay attention to find specific signal characteristics (e.g., K-complexes and spindles), and sometimes need to integrate information from preceding and subsequent epochs in order to make a decision. To imitate this process and to build a more interpretable deep learning model, we propose a neural network based on a convolutional network (CNN) and attention mechanism to perform automatic sleep staging. The CNN learns local signal characteristics, and the attention mechanism excels in learning inter- and intra-epoch features. In experiments on the public sleep-edf and sleep-edfx databases with different training and testing set partitioning methods, our model achieved overall accuracies of 93.7% and 82.8%, and macro-average F1-scores of 84.5 and 77.8, respectively, outperforming recently reported machine learning-based methods.

Joint Classification and Prediction CNN Framework for Automatic Sleep Stage Classification

Correctly identifying sleep stages is important in diagnosing and treating sleep disorders. This paper proposes a joint classification-and-prediction framework based on convolutional neural networks (CNNs) for automatic sleep staging, and, subsequently, introduces a simple yet efficient CNN architecture to power the framework. Given a single input epoch, the novel framework jointly determines its label (classification) and its neighboring epochs' labels (prediction) in the contextual output. While the proposed framework is orthogonal to the widely adopted classification schemes, which take one or multiple epochs as contextual inputs and produce a single classification decision on the target epoch, we demonstrate its advantages in several ways. First, it leverages the dependency among consecutive sleep epochs while surpassing the problems experienced with the common classification schemes. Second, even with a single model, the framework has the capacity to produce multiple decisions, which are essential in obtaining a good performance as in ensemble-of-models methods, with very little induced computational overhead. Probabilistic aggregation techniques are then proposed to leverage the availability of multiple decisions. To illustrate the efficacy of the proposed framework, we conducted experiments on two public datasets: Sleep-EDF Expanded (Sleep-EDF), which consists of 20 subjects, and Montreal Archive of Sleep Studies (MASS) dataset, which consists of 200 subjects. The proposed framework yields an overall classification accuracy of 82.3% and 83.6%, respectively. We also show that the proposed framework not only is superior to the baselines based on the common classification schemes but also outperforms existing deep-learning approaches. To our knowledge, this is the first work going beyond the standard single-output classification to consider multitask neural networks for automatic sleep staging. This framework provides avenues for further studies of different neural-network architectures for automatic sleep staging.

SeqSleepNet: End-to-End Hierarchical Recurrent Neural Network for Sequence-to-Sequence Automatic Sleep Staging

Automatic sleep staging has been often treated as a simple classification problem that aims at determining the label of individual target polysomnography epochs one at a time. In this paper, we tackle the task as a sequence-to-sequence classification problem that receives a sequence of multiple epochs as input and classifies all of their labels at once. For this purpose, we propose a hierarchical recurrent neural network named SeqSleepNet (source code is available at http://github.com/pquochuy/SeqSleepNet). At the epoch processing level, the network consists of a filterbank layer tailored to learn frequency-domain filters for preprocessing and an attention-based recurrent layer designed for short-term sequential modeling. At the sequence processing level, a recurrent layer placed on top of the learned epoch-wise features for long-term modeling of sequential epochs. The classification is then carried out on the output vectors at every time step of the top recurrent layer to produce the sequence of output labels. Despite being hierarchical, we present a strategy to train the network in an end-to-end fashion. We show that the proposed network outperforms the state-of-the-art approaches, achieving an overall accuracy, macro F1-score, and Cohen's kappa of 87.1%, 83.3%, and 0.815 on a publicly available dataset with 200 subjects.