DOSED: A deep learning approach to detect multiple sleep micro-events in EEG signal

Advancing epilepsy diagnosis: A meta-analysis of artificial intelligence approaches for interictal epileptiform discharge detection

INTRODUCTION

Interictal epileptiform discharges (IEDs) in electroencephalograms (EEGs) are an important biomarker for epilepsy. Currently, the gold standard for IED detection is the visual analysis performed by experts. However, this process is expert-biased, and time-consuming. Developing fast, accurate, and robust detection methods for IEDs based on EEG may facilitate epilepsy diagnosis. We aim to assess the performance of deep learning (DL) and classic machine learning (ML) algorithms in classifying EEG segments into IED and non-IED categories, as well as distinguishing whether the entire EEG contains IED or not.

METHODS

We systematically searched PubMed, Embase, and Web of Science following PRISMA guidelines. We excluded studies that only performed the detection of IEDs instead of binary segment classification. Risk of Bias was evaluated with Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2). Meta-analysis with the overall area under the Summary Receiver Operating Characteristic (SROC), sensitivity, and specificity as effect measures, was performed with R software.

RESULTS

A total of 23 studies, comprising 3,629 patients, were eligible for synthesis. Eighteen models performed discharge-level classification, and 6 whole-EEG classification. For the IED-level classification, 3 models were validated in an external dataset with more than 50 patients and achieved a sensitivity of 84.9 % (95 % CI: 82.3-87.2) and a specificity of 68.7 % (95 % CI: 7.9-98.2). Five studies reported model performance using both internal validation (cross-validation) and external datasets. The meta-analysis revealed higher performance for internal validation, with 90.4 % sensitivity and 99.6 % specificity, compared to external validation, which showed 78.1 % sensitivity and 80.1 % specificity.

CONCLUSION

Meta-analysis showed higher performance for models validated with resampling methods compared to those using external datasets. Only a minority of models use more robust validation techniques, which often leads to overfitting.

Unveil sleep spindles with concentration of frequency and time (ConceFT)

Objective.Sleep spindles contain crucial brain dynamics information. We introduce the novel non-linear time-frequency (TF) analysis tool 'Concentration of Frequency and Time' (ConceFT) to create an interpretable automated algorithm for sleep spindle annotation in EEG data and to measure spindle instantaneous frequencies (IFs).Approach.ConceFT effectively reduces stochastic EEG influence, enhancing spindle visibility in the TF representation. Our automated spindle detection algorithm, ConceFT-Spindle (ConceFT-S), is compared to A7 (non-deep learning) and SUMO (deep learning) using Dream and Montreal Archive of Sleep Studies (MASS) benchmark databases. We also quantify spindle IF dynamics.Main results.ConceFT-S achieves F1 scores of 0.765 in Dream and 0.791 in MASS, which surpass A7 and SUMO. We reveal that spindle IF is generally nonlinear.Significance.ConceFT offers an accurate, interpretable EEG-based sleep spindle detection algorithm and enables spindle IF quantification.

Strengths, weaknesses, opportunities, and threats of using AI-enabled technology in sleep medicine: a commentary

Over the past few years, artificial intelligence (AI) has emerged as a powerful tool used to efficiently automate several tasks across multiple domains. Sleep medicine is perfectly positioned to leverage this tool due to the wealth of physiological signals obtained through sleep studies or sleep tracking devices and abundance of accessible clinical data through electronic medical records. However, caution must be applied when utilizing AI, due to intrinsic challenges associated with novel technology. The Artificial Intelligence in Sleep Medicine Committee of the American Academy of Sleep Medicine reviews advancements in AI within the sleep medicine field. In this article, the Artificial Intelligence in Sleep Medicine committee members provide a commentary on the scope of AI technology in sleep medicine. The commentary identifies 3 pivotal areas in sleep medicine that can benefit from AI technologies: clinical care, lifestyle management, and population health management. This article provides a detailed analysis of the strengths, weaknesses, opportunities, and threats associated with using AI-enabled technologies in each pivotal area. Finally, the article broadly reviews barriers and challenges associated with using AI-enabled technologies and offers possible solutions.

CITATION

Bandyopadhyay A, Oks M, Sun H, et al. Strengths, weaknesses, opportunities, and threats of using AI-enabled technology in sleep medicine: a commentary. J Clin Sleep Med. 2024;20(7):1183-1191.

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.

A robust deep learning detector for sleep spindles and K-complexes: towards population norms

Sleep spindles (SSs) and K-complexes (KCs) are brain patterns involved in cognitive functions that appear during sleep. Large-scale sleep studies would benefit from precise and robust automatic sleep event detectors, capable of adapting the variability in both electroencephalography (EEG) signals and expert annotation rules. We introduce the Sleep EEG Event Detector (SEED), a deep learning system that outperforms existing approaches in SS and KC detection, reaching an F1-score of 80.5% and 83.7%, respectively, on the MASS2 dataset. SEED transfers well and requires minimal fine-tuning for new datasets and annotation styles. Remarkably, SEED substantially reduces the required amount of annotated data by using a novel pretraining approach that leverages the rule-based detector A7. An analysis of 11,224 subjects revealed that SEED's detections provide better estimates of SS population statistics than existing approaches. SEED is a powerful resource for obtaining sleep-event statistics that could be useful for establishing population norms.

MSED: A Multi-Modal Sleep Event Detection Model for Clinical Sleep Analysis

Clinical sleep analysis require manual analysis of sleep patterns for correct diagnosis of sleep disorders. However, several studies have shown significant variability in manual scoring of clinically relevant discrete sleep events, such as arousals, leg movements, and sleep disordered breathing (apneas and hypopneas). We investigated whether an automatic method could be used for event detection and if a model trained on all events (joint model) performed better than corresponding event-specific models (single-event models). We trained a deep neural network event detection model on 1653 individual recordings and tested the optimized model on 1000 separate hold-out recordings. F1 scores for the optimized joint detection model were 0.70, 0.63, and 0.62 for arousals, leg movements, and sleep disordered breathing, respectively, compared to 0.65, 0.61, and 0.60 for the optimized single-event models. Index values computed from detected events correlated positively with manual annotations (r2 = 0.73, r2 = 0.77, r2 = 0.78, respectively). We furthermore quantified model accuracy based on temporal difference metrics, which improved overall by using the joint model compared to single-event models. Our automatic model jointly detects arousals, leg movements and sleep disordered breathing events with high correlation with human annotations. Finally, we benchmark against previous state-of-the-art multi-event detection models and found an overall increase in F1 score with our proposed model despite a 97.5% reduction in model size.

Sleep spindles, stress and PTSD: The state of the science and future directions

Sleep spindles are a signature feature of non-REM (NREM) sleep, with demonstrated relationships to sleep maintenance and learning and memory. Because PTSD is characterized by disturbances in sleep maintenance and in stress learning and memory, there is now a growing interest in examining the role of sleep spindles in the neurobiology of PTSD. This review provides an overview of methods for measuring and detecting sleep spindles as they pertain to human PTSD and stress research, presents a critical review of early findings examining sleep spindles in PTSD and stress neurobiology, and proposes several directions for future research. In doing so, this review underscores the extensive heterogeneity in sleep spindle measurement and detection methods, the wide range of spindle features that may be and have been examined, the many persisting unknowns about the clinical and functional relevance of those features, and the problems considering PTSD as a homogeneous group in between-group comparisons. This review also highlights the progress that has been made in this field and underscores the strong rationale for ongoing work in this area.

Detection of K-complexes in EEG waveform images using faster R-CNN and deep transfer learning

Background

The electroencephalography (EEG) signal carries important information about the electrical activity of the brain, which may reveal many pathologies. This information is carried in certain waveforms and events, one of which is the K-complex. It is used by neurologists to diagnose neurophysiologic and cognitive disorders as well as sleep studies. Existing detection methods largely depend on tedious, time-consuming, and error-prone manual inspection of the EEG waveform.

Methods

In this paper, a highly accurate K-complex detection system is developed. Based on multiple convolutional neural network (CNN) feature extraction backbones and EEG waveform images, a regions with faster regions with convolutional neural networks (Faster R-CNN) detector was designed, trained, and tested. Extensive performance evaluation was performed using four deep transfer learning feature extraction models (AlexNet, ResNet-101, VGG19 and Inceptionv3). The dataset was comprised of 10948 images of EEG waveforms, with the location of the K-complexes included as separate text files containing the bounding boxes information.

Results

The Inceptionv3 and VGG19-based detectors performed consistently high (i.e., up to 99.8% precision and 0.2% miss rate) over different testing scenarios, in which the number of training images was varied from 60% to 80% and the positive overlap threshold was increased from 60% to 90%.

Conclusions

Our automated method appears to be a highly accurate automatic K-complex detection in real-time that can aid practitioners in speedy EEG inspection.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12911-022-02042-x.

Age estimation from sleep studies using deep learning predicts life expectancy

Sleep disturbances increase with age and are predictors of mortality. Here, we present deep neural networks that estimate age and mortality risk through polysomnograms (PSGs). Aging was modeled using 2500 PSGs and tested in 10,699 PSGs from men and women in seven different cohorts aged between 20 and 90. Ages were estimated with a mean absolute error of 5.8 ± 1.6 years, while basic sleep scoring measures had an error of 14.9 ± 6.29 years. After controlling for demographics, sleep, and health covariates, each 10-year increment in age estimate error (AEE) was associated with increased all-cause mortality rate of 29% (95% confidence interval: 20-39%). An increase from -10 to +10 years in AEE translates to an estimated decreased life expectancy of 8.7 years (95% confidence interval: 6.1-11.4 years). Greater AEE was mostly reflected in increased sleep fragmentation, suggesting this is an important biomarker of future health independent of sleep apnea.

A Hybrid Expert System for Individualized Quantification of Electrical Status Epilepticus During Sleep Using Biogeography-Based Optimization

Electrical status epilepticus during sleep (ESES) is an epileptic encephalopathy in children with complex clinical manifestations. It is accompanied by specific electroencephalography (EEG) patterns of continuous spike and slow-waves. Quantifying such EEG patterns is critical to the diagnosis of ESES. While most of the existing automatic ESES quantification systems ignore the morphological variations of the signal as well as the individual variability among subjects. To address these issues, this paper presents a hybrid expert system that dedicates to mimicking the decision-making process of clinicians in ESES quantification by taking the morphological variations, individual variability, and medical knowledge into consideration. The proposed hybrid system not only offers a general scheme that could propel a semi-auto morphology analysis-based expert decision model to a fully automated ESES quantification with biogeography-based optimization (BBO), but also proposes a more precise individualized quantification system to involve the personalized characteristics by adopting an individualized parameters-selection framework. The feasibility and reliability of the proposed method are evaluated on a clinical dataset collected from twenty subjects at Children's Hospital of Fudan University, Shanghai, China. The estimation error for the individualized quantitative descriptor ESES is 0-4.32% and the average estimation error is 0.95% for all subjects. Experimental results show the presented system outperforms existing works and the individualized system significantly improves the performance of ESES quantification. The favorable results indicate that the proposed hybrid expert system for automatic ESES quantification is promising to support the diagnosis of ESES.

A Model Visualization-based Approach for Insight into Waveforms and Spectra Learned by CNNs

Recent years have shown a growth in the application of deep learning architectures such as convolutional neural networks (CNNs), to electrophysiology analysis. However, using neural networks with raw time-series data makes explainability a significant challenge. Multiple explainability approaches have been developed for insight into the spectral features learned by CNNs from EEG. However, across electrophysiology modalities, and even within EEG, there are many unique waveforms of clinical relevance. Existing methods that provide insight into waveforms learned by CNNs are of questionable utility. In this study, we present a novel model visualization-based approach that analyzes the filters in the first convolutional layer of the network. To our knowledge, this is the first method focused on extracting explainable information from EEG waveforms learned by CNNs while also providing insight into the learned spectral features. We demonstrate the viability of our approach within the context of automated sleep stage classification, a well-characterized domain that can help validate our approach. We identify 3 subgroups of filters with distinct spectral properties, determine the relative importance of each group of filters, and identify several unique waveforms learned by the classifier that were vital to the classifier performance. Our approach represents a significant step forward in explainability for electrophysiology classifiers, which we also hope will be useful for providing insights in future studies. Clinical Relevance- Our approach can assist with the development and validation of clinical time-series classifiers.

[Sleep spindles-Function, detection and use as biomarker for diagnostics in psychiatry]

BACKGROUND

The sleep spindle is a graphoelement of an electroencephalogram (EEG), which can be observed in light and deep sleep. Alterations in spindle activity have been described for a range of psychiatric disorders. Due to their relatively constant properties, sleep spindles may therefore be potential biomarkers in psychiatric diagnostics.

METHOD

This article presents an overview of the state of the science on the characteristics and functions of the sleep spindle as well as known alterations of spindle activity in psychiatric disorders. Various methodological approaches and developments of spindle detection are explained with respect to their potential for application in psychiatric diagnostics.

RESULTS AND CONCLUSION

Although alterations in spindle activity in psychiatric disorders are known and have been described in detail, their exact potential for psychiatric diagnostics has yet to be fully determined. In this respect, the acquisition of knowledge in research is currently constrained by manual and automated methods for spindle detection, which require high levels of resources and are error prone. Newer approaches to spindle detection based on deep-learning procedures could overcome the difficulties of previous detection methods, and thus open up new possibilities for the practical application of sleep spindles as biomarkers in the psychiatric practice.

Advanced sleep spindle identification with neural networks

Sleep spindles are neurophysiological phenomena that appear to be linked to memory formation and other functions of the central nervous system, and that can be observed in electroencephalographic recordings (EEG) during sleep. Manually identified spindle annotations in EEG recordings suffer from substantial intra- and inter-rater variability, even if raters have been highly trained, which reduces the reliability of spindle measures as a research and diagnostic tool. The Massive Online Data Annotation (MODA) project has recently addressed this problem by forming a consensus from multiple such rating experts, thus providing a corpus of spindle annotations of enhanced quality. Based on this dataset, we present a U-Net-type deep neural network model to automatically detect sleep spindles. Our model’s performance exceeds that of the state-of-the-art detector and of most experts in the MODA dataset. We observed improved detection accuracy in subjects of all ages, including older individuals whose spindles are particularly challenging to detect reliably. Our results underline the potential of automated methods to do repetitive cumbersome tasks with super-human performance.

Early Stroke Prediction Methods for Prevention of Strokes

The emergence of the latest technologies gives rise to the usage of noninvasive techniques for assisting health-care systems. Amongst the four major cardiovascular diseases, stroke is one of the most dangerous and life-threatening disease, but the life of a patient can be saved if the stroke is detected during early stage. The literature reveals that the patients always experience ministrokes which are also known as transient ischemic attacks (TIA) before experiencing the actual attack of the stroke. Most of the literature work is based on the MRI and CT scan images for classifying the cardiovascular diseases including a stroke which is an expensive approach for diagnosis of early strokes. In India where cases of strokes are rising, there is a need to explore noninvasive cheap methods for the diagnosis of early strokes. Hence, this problem has motivated us to conduct the study presented in this paper. A noninvasive approach for the early diagnosis of the strokes is proposed. The cascaded prediction algorithms are time-consuming in producing the results and cannot work on the raw data and without making use of the properties of EEG. Therefore, the objective of this paper is to devise mechanisms to forecast strokes on the basis of processed EEG data. This paper is proposing time series-based approaches such as LSTM, biLSTM, GRU, and FFNN that can handle time series-based predictions to make useful decisions. The experimental research outcome reveals that all the algorithms taken up for the research study perform well on the prediction problem of early stroke detection, but GRU performs the best with 95.6% accuracy, whereas biLSTM gives 91% accuracy and LSTM gives 87% accuracy and FFNN gives 83% accuracy. The experimental outcome is able to measure the brain waves to predict the signs of strokes. The findings can certainly assist the physicians to detect the stroke at early stages to save the lives of the patients.

SpindleU-Net: An Adaptive U-Net Framework for Sleep Spindle Detection in Single-Channel EEG

The sleep spindles in EEG have become one type of biomarker used to assess cognitive abilities and related disorders, and thus their detection is crucial for clinical research. This task, traditionally performed by sleep experts, is time-consuming. Many methods have been proposed to automate this process, yet an increase in performance is still expected. Inspired by the application in image segmentation, we propose a point-wise spindle detection method based on the U-Net framework with an attention module (SpindleU-Net). It maps the sequences of arbitrary-length EEG inputs to those of dense labels of spindle or non-spindle on freely chosen intervals. The attention module that focuses on the salient spindle region allows better performance, and a task-specific loss function is defined to alleviate the problem of imbalanced classification. As a deep learning method, SpindleU-Net outperforms state-of-the-art methods on the widely used benchmark dataset of MASS as well as the DREAMS dataset with a small number of samples. On MASS dataset it achieves average F1 scores of 0.854 and 0.803 according to its consistency with the annotations by two sleep experts respectively. On DREAMS dataset, it shows the average F1 score of 0.739. Its cross-dataset performance is also better compared to other methods, showing the good generalization ability for cross-dataset applications.

Deep Learning-Based Stroke Disease Prediction System Using Real-Time Bio Signals

The emergence of an aging society is inevitable due to the continued increases in life expectancy and decreases in birth rate. These social changes require new smart healthcare services for use in daily life, and COVID-19 has also led to a contactless trend necessitating more non-face-to-face health services. Due to the improvements that have been achieved in healthcare technologies, an increasing number of studies have attempted to predict and analyze certain diseases in advance. Research on stroke diseases is actively underway, particularly with the aging population. Stroke, which is fatal to the elderly, is a disease that requires continuous medical observation and monitoring, as its recurrence rate and mortality rate are very high. Most studies examining stroke disease to date have used MRI or CT images for simple classification. This clinical approach (imaging) is expensive and time-consuming while requiring bulky equipment. Recently, there has been increasing interest in using non-invasive measurable EEGs to compensate for these shortcomings. However, the prediction algorithms and processing procedures are both time-consuming because the raw data needs to be separated before the specific attributes can be obtained. Therefore, in this paper, we propose a new methodology that allows for the immediate application of deep learning models on raw EEG data without using the frequency properties of EEG. This proposed deep learning-based stroke disease prediction model was developed and trained with data collected from real-time EEG sensors. We implemented and compared different deep-learning models (LSTM, Bidirectional LSTM, CNN-LSTM, and CNN-Bidirectional LSTM) that are specialized in time series data classification and prediction. The experimental results confirmed that the raw EEG data, when wielded by the CNN-bidirectional LSTM model, can predict stroke with 94.0% accuracy with low FPR (6.0%) and FNR (5.7%), thus showing high confidence in our system. These experimental results demonstrate the feasibility of non-invasive methods that can easily measure brain waves alone to predict and monitor stroke diseases in real time during daily life. These findings are expected to lead to significant improvements for early stroke detection with reduced cost and discomfort compared to other measuring techniques.

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.

The Dreem Headband compared to polysomnography for electroencephalographic signal acquisition and sleep staging

Study Objectives

The development of ambulatory technologies capable of monitoring brain activity during sleep longitudinally is critical for advancing sleep science. The aim of this study was to assess the signal acquisition and the performance of the automatic sleep staging algorithms of a reduced-montage dry-electroencephalographic (EEG) device (Dreem headband, DH) compared to the gold-standard polysomnography (PSG) scored by five sleep experts.

Methods

A total of 25 subjects who completed an overnight sleep study at a sleep center while wearing both a PSG and the DH simultaneously have been included in the analysis. We assessed (1) similarity of measured EEG brain waves between the DH and the PSG; (2) the heart rate, breathing frequency, and respiration rate variability (RRV) agreement between the DH and the PSG; and (3) the performance of the DH’s automatic sleep staging according to American Academy of Sleep Medicine guidelines versus PSG sleep experts manual scoring.

Results

The mean percentage error between the EEG signals acquired by the DH and those from the PSG for the monitoring of α was 15 ± 3.5%, 16 ± 4.3% for β, 16 ± 6.1% for λ, and 10 ± 1.4% for θ frequencies during sleep. The mean absolute error for heart rate, breathing frequency, and RRV was 1.2 ± 0.5 bpm, 0.3 ± 0.2 cpm, and 3.2 ± 0.6%, respectively. Automatic sleep staging reached an overall accuracy of 83.5 ± 6.4% (F1 score: 83.8 ± 6.3) for the DH to be compared with an average of 86.4 ± 8.0% (F1 score: 86.3 ± 7.4) for the 5 sleep experts.

Conclusions

These results demonstrate the capacity of the DH to both monitor sleep-related physiological signals and process them accurately into sleep stages. This device paves the way for, large-scale, longitudinal sleep studies.

Clinical Trial Registration

NCT03725943.

An energy screening and morphology characterization-based hybrid expert scheme for automatic identification of micro-sleep event K-complex

BACKGROUND AND OBJECTIVE

K-complexes, as a significant indicator in sleep staging and sleep protection, are an important micro-event in sleep analysis. Clinically, K-complexes are recognized through the expert visual inspection of electroencephalogram (EEG) during sleep. Since this process is laborious and has high inter-observer variability, developing automated K-complex detection methods can alleviate the burden on clinicians while providing reliable recognition results. However, existing methods face the following issues. First, most work only identifies the K-complexes in stage 2, which requires distinguishing the sleep stages as the prerequisite for further events' identification. Second, most approaches can only detect the occurrence of events without the ability to predict their location and duration, which are also essential to sleep analysis.

METHODS

In this work, a novel hybrid expert scheme for K-complex detection is proposed by integrating signal morphology with expert knowledge into the decision-making process. To eliminate artifacts, and to minimize the individual variability in raw sleep EEG signals, the potential K-complex candidates are first screened by combining Teager energy operator (TEO) and personalized thresholds. Then, to distinguish signal shapes from background activity, a novel frame of filtering based on morphological filtering (MF) is devised to differentiate morphological components of K-complex waveforms from EEG series. Finally, K-complex waveforms are identified from the extracted morphological information by judgment rules, which are inspired by expert knowledge of micro-sleep events.

RESULTS

Detection performance is evaluated by its application on the public database MASS-C1 (Montreal archives of sleep studies cohort one) which includes the recordings of 19 healthy adults. The detection performance demonstrates an F-measure of 0.63 with a recall of 0.81 and a precision of 0.53 on average. The duration error between events and detections is 0.10 s.

CONCLUSIONS

The presented scheme has detected the occurrence of events. Meanwhile, it has recognized their locations and durations. The favorable results exhibit that the proposed scheme outperforms the state-of-the-art studies and has great potential to help release the burden of experts in sleep EEG analysis.

Deep learning for robust detection of interictal epileptiform discharges

OBJECTIVE

Automatic detection of interictal epileptiform discharges (IEDs, short as ``spikes'') from an epileptic brain can help predict seizure recurrence and support the diagnosis of epilepsy. Developing fast, reliable and robust detection methods for IEDs based on scalp or intracortical EEG may facilitate online seizure monitoring and closed-loop neurostimulation.

APPROACH

We developed a new deep learning approach, which employs a long short-term memory (LSTM) network architecture (``IEDnet'') and an auxiliary classifier generative adversarial network (AC-GAN), to train on both expert-annotated and augmented spike events from intracranial electroencephalography (iEEG) recordings of epilepsy patients. We validated our IEDnet with two real-world iEEG datasets, and compared IEDnet with the support vector machine (SVM) and random forest (RF) classifiers on their detection performances.

MAIN RESULTS

IEDnet achieved excellent cross-validated detection performances in terms of both sensitivity and specificity, and outperformed SVM and RF. Synthetic spike samples augmented by AC-GAN further improved the detection performance. In addition, the performance of IEDnet was robust with respect to the sampling frequency and noise. Furthermore, we also demonstrated the cross-institutional generalization ability of IEDnet while testing between two datasets.

SIGNIFICANCE

IEDnet achieves excellent detection performances in identifying interictal spikes. AC-GAN can produce augmented iEEG samples to improve supervised deep learning.

Machine-Learning-Based Elderly Stroke Monitoring System Using Electroencephalography Vital Signals

Stroke is the third highest cause of death worldwide after cancer and heart disease, and the number of stroke diseases due to aging is set to at least triple by 2030. As the top three causes of death worldwide are all related to chronic disease, the importance of healthcare is increasing even more. Models that can predict real-time health conditions and diseases using various healthcare services are attracting increasing attention. Most diagnosis and prediction methods of stroke for the elderly involve imaging techniques such as magnetic resonance imaging (MRI). It is difficult to rapidly and accurately diagnose and predict stroke diseases due to the long testing times and high costs associated with MRI. Thus, in this paper, we design and implement a health monitoring system that can predict the precursors of stroke diseases in the elderly in real time during daily walking. First, raw electroencephalography (EEG) data from six channels were preprocessed via Fast Fourier Transform (FFT). The raw EEG power values were then extracted from the raw spectra: alpha (α), beta (β), gamma (γ), delta (δ), and theta (θ) as well as the low β, high β, and θ to β ratio, respectively. The experiments in this paper confirm that the important features of EEG biometric signals alone during walking can accurately determine stroke precursors and occurrence in the elderly with more than 90% accuracy. Further, the Random Forest algorithm with quartiles and Z-score normalization validates the clinical significance and performance of the system proposed in this paper with a 92.51% stroke prediction accuracy. The proposed system can be implemented at a low cost, and it can be applied for early disease detection and prediction using the precursor symptoms of real-time stroke. Furthermore, it is expected that it will be able to detect other diseases such as cancer and heart disease in the future.

Deep transfer learning for improving single-EEG arousal detection

Datasets in sleep science present challenges for machine learning algorithms due to differences in recording setups across clinics. We investigate two deep transfer learning strategies for overcoming the channel mismatch problem for cases where two datasets do not contain exactly the same setup leading to degraded performance in single-EEG models. Specifically, we train a baseline model on multivariate polysomnography data and subsequently replace the first two layers to prepare the architecture for single-channel electroencephalography data. Using a fine-tuning strategy, our model yields similar performance to the baseline model (F1=0.682 and F1=0.694, respectively), and was significantly better than a comparable single-channel model. Our results are promising for researchers working with small databases who wish to use deep learning models pre-trained on larger databases.

Beyond K-complex binary scoring during sleep: probabilistic classification using deep learning

STUDY OBJECTIVES

K-complexes (KCs) are a recognized electroencephalography marker of sensory processing and a defining feature of sleep stage 2. KC frequency and morphology may also be reflective of sleep quality, aging, and a range of sleep and sensory processing deficits. However, manual scoring of K-complexes is impractical, time-consuming, and thus costly and currently not well-standardized. Although automated KC detection methods have been developed, performance and uptake remain limited.

METHODS

The proposed algorithm is based on a deep neural network and Gaussian process, which gives the input waveform a probability of being a KC ranging from 0% to 100%. The algorithm was trained on half a million synthetic KCs derived from manually scored sleep stage 2 KCs from the Montreal Archive of Sleep Study containing 19 healthy young participants. Algorithm performance was subsequently assessed on 700 independent recordings from the Cleveland Family Study using sleep stages 2 and 3 data.

RESULTS

The developed algorithm showed an F1 score (a measure of binary classification accuracy) of 0.78 and thus outperforms currently available KC scoring algorithms with F1 = 0.2-0.6. The probabilistic approach also captured expected variability in KC shape and amplitude within individuals and across age groups.

CONCLUSIONS

An automated probabilistic KC classification is well suited and effective for systematic KC detection for a more in-depth exploration of potential relationships between KCs during sleep and clinical outcomes such as health impacts and daytime symptomatology.

Automatic detection of pulmonary nodules on CT images with YOLOv3: development and evaluation using simulated and patient data

BACKGROUND

To develop a high-efficiency pulmonary nodule computer-aided detection (CAD) method for localization and diameter estimation.

METHODS

The developed CAD method centralizes a novel convolutional neural network (CNN) algorithm, You Only Look Once (YOLO) v3, as a deep learning approach. This method is featured by two distinct properties: (I) an automatic multi-scale feature extractor for nodule feature screening, and (II) a feature-based bounding box generator for nodule localization and diameter estimation. Two independent studies were performed to train and evaluate this CAD method. One study comprised of a computer simulation that utilized computer-based ground truth. In this study, 300 CT scans were simulated by Cardiac-torso (XCAT) digital phantom. Spherical nodules of various sizes (i.e., 3-10 mm in diameter) were randomly implanted within the lung region of the simulated images-the second study utilized human-based ground truth in patients. The CAD method was developed by CT scans sourced from the LIDC-IDRI database. CT scans with slice thickness above 2.5 mm were excluded, leaving 888 CT images for analysis. A 10-fold cross-validation procedure was implemented in both studies to evaluate network hyper-parameterization and generalization. The overall accuracy of the CAD method was evaluated by the detection sensitivities, in response to average false positives (FPs) per image. In the patient study, the detection accuracy was further compared against 9 recently published CAD studies using free-receiver response operating characteristic (FROC) curve analysis. Localization and diameter estimation accuracies were quantified by the mean and standard error between the predicted value and ground truth.

RESULTS

The average results among the 10 cross-validation folds in both studies demonstrated the CAD method achieved high detection accuracy. The sensitivity was 99.3% (FPs =1), and improved to 100% (FPs =4) in the simulation study. The corresponding sensitivities were 90.0% and 95.4% in the patient study, displaying superiority over several conventional and CNN-based lung nodule CAD methods in the FROC curve analysis. Nodule localization and diameter estimation errors were less than 1 mm in both studies. The developed CAD method achieved high computational efficiency: it yields nodule-specific quantitative values (i.e., number, existence confidence, central coordinates, and diameter) within 0.1 s for 2D CT slice inputs.

CONCLUSIONS

The reported results suggest that the developed lung pulmonary nodule CAD method possesses high accuracies of nodule localization and diameter estimation. The high computational efficiency enables its potential clinical application in the future.

Deep Learning in Physiological Signal Data: A Survey

Deep Learning (DL), a successful promising approach for discriminative and generative tasks, has recently proved its high potential in 2D medical imaging analysis; however, physiological data in the form of 1D signals have yet to be beneficially exploited from this novel approach to fulfil the desired medical tasks. Therefore, in this paper we survey the latest scientific research on deep learning in physiological signal data such as electromyogram (EMG), electrocardiogram (ECG), electroencephalogram (EEG), and electrooculogram (EOG). We found 147 papers published between January 2018 and October 2019 inclusive from various journals and publishers. The objective of this paper is to conduct a detailed study to comprehend, categorize, and compare the key parameters of the deep-learning approaches that have been used in physiological signal analysis for various medical applications. The key parameters of deep-learning approach that we review are the input data type, deep-learning task, deep-learning model, training architecture, and dataset sources. Those are the main key parameters that affect system performance. We taxonomize the research works using deep-learning method in physiological signal analysis based on: (1) physiological signal data perspective, such as data modality and medical application; and (2) deep-learning concept perspective such as training architecture and dataset sources.

Integrating sleep, neuroimaging, and computational approaches for precision psychiatry

In advancing precision psychiatry, we focus on what imaging technology and computational approaches offer for the future of diagnostic subtyping and personalized tailoring of interventions for sleep impairment in mood and anxiety disorders. Current diagnostic criteria for mood and anxiety tend to lump different forms of sleep disturbance together. Parsing the biological features of sleep impairment and brain circuit dysfunction is one approach to identifying subtypes within these disorders that are mechanistically coherent and offer targets for intervention. We focus on two large-scale neural circuits implicated in sleep impairment and in mood and anxiety disorders: the default mode network and negative affective network. Through a synthesis of existing knowledge about these networks, we pose a testable framework for understanding how hyper- versus hypo-engagement of these networks may underlie distinct features of mood and sleep impairment. Within this framework we consider whether poor sleep quality may have an explanatory role in previously observed associations between network dysfunction and mood symptoms. We expand this framework to future directions including the potential for connecting circuit-defined subtypes to more distal features derived from digital phenotyping and wearable technologies, and how new discovery may be advanced through machine learning approaches.

Towards a Flexible Deep Learning Method for Automatic Detection of Clinically Relevant Multi-Modal Events in the Polysomnogram

Much attention has been given to automatic sleep staging algorithms in past years, but the detection of discrete events in sleep studies is also crucial for precise characterization of sleep patterns and possible diagnosis of sleep disorders. We propose here a deep learning model for automatic detection and annotation of arousals and leg movements. Both of these are commonly seen during normal sleep, while an excessive amount of either is linked to disrupted sleep patterns, excessive daytime sleepiness impacting quality of life, and various sleep disorders. Our model was trained on 1,485 subjects and tested on 1,000 separate recordings of sleep. We tested two different experimental setups and found optimal arousal detection was attained by including a recurrent neural network module in our default model with a dynamic default event window (F1 = 0.75), while optimal leg movement detection was attained using a static event window (F1 = 0.65). Our work show promise while still allowing for improvements. Specifically, future research will explore the proposed model as a general-purpose sleep analysis model.

Advanced polysomnographic analysis for OSA: A pathway to personalized management?

Obstructive sleep apnea (OSA) is a highly heterogeneous disorder, with diverse pathways to disease, expression of disease, susceptibility to co-morbidities and response to therapy, and is ideally suited to precision medicine approaches. Clinically, the content of the information-rich polysomnogram (PSG) is not currently fully utilized in determining patient management. Novel PSG parameters such as hypoxic burden, pulse transit time, cardiopulmonary coupling and the frequency representations of PSG sensor signals could predict a variety of cardiovascular disease, cancer and neurodegeneration co-morbidities. The PSG can also be used to identify key pathophysiological parameters such as loop gain, arousal threshold and muscle compensation which can enhance understanding of the causes of OSA in an individual, and thereby guide choices on therapy. Machine learning methods performing their own parameter extraction coupled with large PSG data sets offer an exciting opportunity for discovering new links between the PSG variables and disease outcomes. By exploiting existing and emerging analytical methods, the PSG may offer a pathway to personalized management for OSA.