DiBa: a data-driven Bayesian algorithm for sleep spindle detection

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.

Deep-spindle: An automated sleep spindle detection system for analysis of infant sleep spindles

BACKGROUND

Sleep spindles are an indicator of the development and integrity of the central nervous system in infants. Identifying sleep spindles manually in EEG is time-consuming and typically requires experienced experts. Automated detection of sleep spindles would greatly facilitate this analysis. Deep learning methods have been widely used recently in EEG analysis.

METHOD

We have developed a deep learning-based automated sleep spindle detection system, Deep-spindle, which employs a convolutional neural network (CNN) combined with a bidirectional Long Short-Term Memory (LSTM) network, which could assist in the analysis of infant sleep spindles. Deep-spindle was trained on the EEGs of ex-term infants to estimate the number and duration of sleep spindles. The ex-term EEG on channel F4-C4 was split into training (N=81) and validation (N=30) sets. An additional 30 ex-term EEG and 54 ex-preterm infant EEGs (channel F4-C4 and F3-C3) were used as an independent test set.

RESULT

Deep-spindle detected the number of sleep spindles with 91.9% to 96.5% sensitivity and 95.3% to 96.7% specificity, and estimated sleep spindle duration with a percent error of 13.1% to 19.1% in the independent test set. For each detected spindle event, the user is presented with amplitude, power spectral density and the spectrogram of the corresponding spindle EEG, and the probability of the event being a sleep spindle event, providing the user with insight into why the event is predicted as a sleep spindle to provide confidence in the predictions.

CONCLUSION

The Deep-spindle system can reduce physicians' workload, demonstrating the potential to assist physicians in the automated analysis of sleep spindles in infants.

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.

Spindle-AI: Sleep spindle number and duration estimation in infant EEG

OBJECTIVE

Sleep spindle features show developmental changes during infancy and have the potential to provide an early biomarker for abnormal brain maturation. Manual identification of sleep spindles in the electroencephalogram (EEG) is time-consuming and typically requires highly-trained experts. Automated detection of sleep spindles would greatly facilitate this analysis. Research on the automatic detection of sleep spindles in infant EEG has been limited to-date.

METHODS

We present a random forest-based sleep spindle detection method (Spindle-AI) to estimate the number and duration of sleep spindles in EEG collected from 141 ex-term born infants, recorded at 4 months of age. The signal on channel F4-C4 was split into a training set (81 ex-term) and a validation set (30 ex-term). An additional 30 ex-term infant EEGs (channel F4-C4 and channel F3-C3) were used as an independent test set. Fourteen features were selected for input into a random forest algorithm to estimate the number and duration of spindles and the results were compared against sleep spindles annotated by an experienced clinical physiologist.

RESULTS

The prediction of the number of sleep spindles in the independent test set demonstrated 93.3% to 93.9% sensitivity, 90.7% to 91.5% specificity, and 89.2% to 90.1% precision. The duration estimation of sleep spindle events in the independent test set showed a percent error of 5.7% to 7.4%.

CONCLUSION AND SIGNIFICANCE

Spindle-AI has been implemented as a web server that has the potential to assist clinicians in the fast and accurate monitoring of sleep spindles in infant EEGs.

Chaos in memory function of sleep: A nonlinear dynamical analysis in thalamocortical study

Studying the dynamical behaviors of neuronal models may help in better understanding of real nervous system. In addition, it can help researchers to understand some specific phenomena in neuronal system. The thalamocortical network is made of neurons in the thalamus and cortex. In it, the memory function is consolidated in sleep by creating up and down state oscillations (1 Hz) and fast (13-17 Hz) - slow (8-12 Hz) spindles. Recently, a nonlinear biological model for up-down oscillations and fast-slow spindles of the thalamocortical network has been proposed. In this research, the power spectral for the fast-slow spindle of the model is extracted. Dynamical properties of the model, such as the bifurcation diagrams, and attractors are investigated. The results show that the variation of the synaptic power between the excitatory neurons of the cortex and the reticular neurons in the thalamus changes the spindles' activity. According to previous experimental findings, it is an essential rule for consolidating the memory function during sleep. It is also pointed out that when the fast-slow spindles of the brain increase, the dynamics of the thalamocortical system tend to chaos.

Sleep Spindle Detection Using RUSBoost and Synchrosqueezed Wavelet Transform

Sleep spindles are important electroencephalographic (EEG) waveforms in sleep medicine; however, it is burdensome even for experts to detect spindles, so automatic spindle detection methodologies have been investigated. Conventional methods utilize waveforms template matching or machine learning for detecting spindles. In the former approach, it is necessary to tune thresholds for individual adaptation, while the latter approach has the problem of imbalanced data because the amount of sleep spindles is small compared with the entire EEG data. The present work proposes a sleep spindle detection method that combines wavelet synchrosqueezed transform (SST) and random under-sampling boosting (RUSBoost). SST is a time-frequency analysis method suitable for extracting features of spindle waveforms. RUSBoost is a framework for coping with the imbalanced data problem. The proposed SST-RUS can deal with the imbalanced data in spindle detection and does not require threshold tuning because RUSBoost uses majority voting of weak classifiers for discrimination. The performance of SST-RUS was validated using an open-access database called the Montreal archives of sleep studies cohort 1 (MASS-C1), which showed an F-measure of 0.70 with a sensitivity of 76.9% and a positive predictive value of 61.2%. The proposed method can reduce the burden of PSG scoring.

A deep learning approach for real-time detection of sleep spindles

OBJECTIVE

Sleep spindles have been implicated in memory consolidation and synaptic plasticity during NREM sleep. Detection accuracy and latency in automatic spindle detection are critical for real-time applications.

APPROACH

Here we propose a novel deep learning strategy (SpindleNet) to detect sleep spindles based on a single EEG channel. While the majority of spindle detection methods are used for off-line applications, our method is well suited for online applications.

MAIN RESULTS

Compared with other spindle detection methods, SpindleNet achieves superior detection accuracy and speed, as demonstrated in two publicly available expert-validated EEG sleep spindle datasets. Our real-time detection of spindle onset achieves detection latencies of 150-350 ms (~two-three spindle cycles) and retains excellent performance under low EEG sampling frequencies and low signal-to-noise ratios. SpindleNet has good generalization across different sleep datasets from various subject groups of different ages and species.

SIGNIFICANCE

SpindleNet is ultra-fast and scalable to multichannel EEG recordings, with an accuracy level comparable to human experts, making it appealing for long-term sleep monitoring and closed-loop neuroscience experiments.

Quantitative Analysis of a Marked Point Process based Sleep Spindle Detector (MPP-SSD)

Sleep spindles result from interactions between the thalamic and cortical neurons during the NREM2 stage. Studies show that these waxing and waning episodes of field potentials may have an implied role in memory consolidation, cellular plasticity and neuronal development besides serving as important markers in several neuronal pathologies. For these reasons, accurate spindle scoring of polysomnographic signals is important and has garnered interest in automating the tedious process of scoring via visual inspection. In this paper, we employ a transient model for automatic sleep spindle detection designed as a Marked Point Process (MPP). Further, in order to simplify the model development, the determination of the atoms was done independently for each of the EEG bands. However, this brings the problem of quantifying the effect of the required bandpass filtering, which was not done in previous work. Here we change the Q- factor of the filters and evaluate the effect on the detections provided by the model, when compared with two sleep experts. Several statistics are utilized, and we conclude that the design of the bandpass filters affects the performance. Low Q filters were thought necessary, but the results show that the optimal Q - factor is around 2.

Multichannel sleep spindle detection using sparse low-rank optimization

BACKGROUND

Automated single-channel spindle detectors, for human sleep EEG, are blind to the presence of spindles in other recorded channels unlike visual annotation by a human expert.

NEW METHOD

We propose a multichannel spindle detection method that aims to detect global and local spindle activity in human sleep EEG. Using a non-linear signal model, which assumes the input EEG to be the sum of a transient and an oscillatory component, we propose a multichannel transient separation algorithm. Consecutive overlapping blocks of the multichannel oscillatory component are assumed to be low-rank whereas the transient component is assumed to be piecewise constant with a zero baseline. The estimated oscillatory component is used in conjunction with a bandpass filter and the Teager operator for detecting sleep spindles.

RESULTS AND COMPARISON WITH OTHER METHODS

The proposed method is applied to two publicly available databases and compared with 7 existing single-channel automated detectors. F₁ scores for the proposed spindle detection method averaged 0.66 (0.02) and 0.62 (0.06) for the two databases, respectively. For an overnight 6 channel EEG signal, the proposed algorithm takes about 4min to detect sleep spindles simultaneously across all channels with a single setting of corresponding algorithmic parameters.

CONCLUSIONS

The proposed method attempts to mimic and utilize, for better spindle detection, a particular human expert behavior where the decision to mark a spindle event may be subconsciously influenced by the presence of a spindle in EEG channels other than the central channel visible on a digital screen.

Fast and Stable Signal Deconvolution via Compressible State-Space Models

OBJECTIVE

Common biological measurements are in the form of noisy convolutions of signals of interest with possibly unknown and transient blurring kernels. Examples include EEG and calcium imaging data. Thus, signal deconvolution of these measurements is crucial in understanding the underlying biological processes. The objective of this paper is to develop fast and stable solutions for signal deconvolution from noisy, blurred, and undersampled data, where the signals are in the form of discrete events distributed in time and space.

METHODS

We introduce compressible state-space models as a framework to model and estimate such discrete events. These state-space models admit abrupt changes in the states and have a convergent transition matrix, and are coupled with compressive linear measurements. We consider a dynamic compressive sensing optimization problem and develop a fast solution, using two nested expectation maximization algorithms, to jointly estimate the states as well as their transition matrices. Under suitable sparsity assumptions on the dynamics, we prove optimal stability guarantees for the recovery of the states and present a method for the identification of the underlying discrete events with precise confidence bounds.

RESULTS

We present simulation studies as well as application to calcium deconvolution and sleep spindle detection, which verify our theoretical results and show significant improvement over existing techniques.

CONCLUSION

Our results show that by explicitly modeling the dynamics of the underlying signals, it is possible to construct signal deconvolution solutions that are scalable, statistically robust, and achieve high temporal resolution.

SIGNIFICANCE

Our proposed methodology provides a framework for modeling and deconvolution of noisy, blurred, and undersampled measurements in a fast and stable fashion, with potential application to a wide range of biological data.

Sleep Spindles as an Electrographic Element: Description and Automatic Detection Methods

Sleep spindle is a peculiar oscillatory brain pattern which has been associated with a number of sleep (isolation from exteroceptive stimuli, memory consolidation) and individual characteristics (intellectual quotient). Oddly enough, the definition of a spindle is both incomplete and restrictive. In consequence, there is no consensus about how to detect spindles. Visual scoring is cumbersome and user dependent. To analyze spindle activity in a more robust way, automatic sleep spindle detection methods are essential. Various algorithms were developed, depending on individual research interest, which hampers direct comparisons and meta-analyses. In this review, sleep spindle is first defined physically and topographically. From this general description, we tentatively extract the main characteristics to be detected and analyzed. A nonexhaustive list of automatic spindle detection methods is provided along with a description of their main processing principles. Finally, we propose a technique to assess the detection methods in a robust and comparable way.

Automatic sleep spindle detection: benchmarking with fine temporal resolution using open science tools

Sleep spindle properties index cognitive faculties such as memory consolidation and diseases such as major depression. For this reason, scoring sleep spindle properties in polysomnographic recordings has become an important activity in both research and clinical settings. The tediousness of this manual task has motivated efforts for its automation. Although some progress has been made, increasing the temporal accuracy of spindle scoring and improving the performance assessment methodology are two aspects needing more attention. In this paper, four open-access automated spindle detectors with fine temporal resolution are proposed and tested against expert scoring of two proprietary and two open-access databases. Results highlight several findings: (1) that expert scoring and polysomnographic databases are important confounders when comparing the performance of spindle detectors tested using different databases or scorings; (2) because spindles are sparse events, specificity estimates are potentially misleading for assessing automated detector performance; (3) reporting the performance of spindle detectors exclusively with sensitivity and specificity estimates, as is often seen in the literature, is insufficient; including sensitivity, precision and a more comprehensive statistic such as Matthew's correlation coefficient, F1-score, or Cohen's κ is necessary for adequate evaluation; (4) reporting statistics for some reasonable range of decision thresholds provides a much more complete and useful benchmarking; (5) performance differences between tested automated detectors were found to be similar to those between available expert scorings; (6) much more development is needed to effectively compare the performance of spindle detectors developed by different research teams. Finally, this work clarifies a long-standing but only seldomly posed question regarding whether expert scoring truly is a reliable gold standard for sleep spindle assessment.

Abnormal dynamics of EEG oscillations in schizophrenia patients on multiple time scales

Neuronal oscillations reflect the activity of neuronal ensembles engaged in integrative cognition, and may serve as a functional measure for the cognitive impairment in schizophrenia. This study aims to reveal the abnormal amplitude dynamics of electroencephalogram (EEG) oscillations in schizophrenia patients on multiple time scales. EEGs were recorded from schizophrenia patients ( n = 19) and healthy controls ( n = 16) while they were at resting state with eyes closed, at resting state with eyes open, and at watching video. Detrended fluctuation analysis and measures of life-time and waiting-time were used to characterize the abnormal dynamics of EEG oscillations on both long (1-20 s) and short (≤1 s) time scales. Abnormal dynamics of EEG oscillations in alpha and beta bands were observed. In particular, compared with healthy controls, schizophrenia patients have smaller DFA exponent (implying weaker long-range temporal correlation) in the left fronto-temporal area and smaller DFA exponent, smaller life-time (indicating shorter oscillation burst), and smaller waiting-time in the occipital area in beta band at resting state with eyes open. In addition, schizophrenia patients have larger DFA exponent, larger life-time, and larger waiting-time at some clustered channels in the temporo-parietal area in alpha band at watching video. The present results provide new insights for cognitive deficits and the underlying neuronal dysfunction in schizophrenia.

Combining time-frequency and spatial information for the detection of sleep spindles

EEG sleep spindles are short (0.5-2.0 s) bursts of activity in the 11-16 Hz band occurring during non-rapid eye movement (NREM) sleep. This sporadic activity is thought to play a role in memory consolidation, brain plasticity, and protection of sleep integrity. Many automatic detectors have been proposed to assist or replace experts for sleep spindle scoring. However, these algorithms usually detect too many events making it difficult to achieve a good tradeoff between sensitivity (Se) and false detection rate (FDr). In this work, we propose a semi-automatic detector comprising a sensitivity phase based on well-established criteria followed by a specificity phase using spatial and spectral criteria. In the sensitivity phase, selected events are those which amplitude in the 10-16 Hz band and spectral ratio characteristics both reject a null hypothesis (p < 0.1) stating that the considered event is not a spindle. This null hypothesis is constructed from events occurring during rapid eye movement (REM) sleep epochs. In the specificity phase, a hierarchical clustering of the selected candidates is done based on events' frequency and spatial position along the anterior-posterior axis. Only events from the classes grouping most (at least 80%) spindles scored by an expert are kept. We obtain Se = 93.2% and FDr = 93.0% in the first phase and Se = 85.4% and FDr = 86.2% in the second phase. For these two phases, Matthew's correlation coefficients are respectively 0.228 and 0.324. Results suggest that spindles are defined by specific spatio-spectral properties and that automatic detection methods can be improved by considering these features.

Learning machines and sleeping brains: Automatic sleep stage classification using decision-tree multi-class support vector machines

BACKGROUND

Sleep staging is a critical step in a range of electrophysiological signal processing pipelines used in clinical routine as well as in sleep research. Although the results currently achievable with automatic sleep staging methods are promising, there is need for improvement, especially given the time-consuming and tedious nature of visual sleep scoring.

NEW METHOD

Here we propose a sleep staging framework that consists of a multi-class support vector machine (SVM) classification based on a decision tree approach. The performance of the method was evaluated using polysomnographic data from 15 subjects (electroencephalogram (EEG), electrooculogram (EOG) and electromyogram (EMG) recordings). The decision tree, or dendrogram, was obtained using a hierarchical clustering technique and a wide range of time and frequency-domain features were extracted. Feature selection was carried out using forward sequential selection and classification was evaluated using k-fold cross-validation.

RESULTS

The dendrogram-based SVM (DSVM) achieved mean specificity, sensitivity and overall accuracy of 0.92, 0.74 and 0.88 respectively, compared to expert visual scoring. Restricting DSVM classification to data where both experts' scoring was consistent (76.73% of the data) led to a mean specificity, sensitivity and overall accuracy of 0.94, 0.82 and 0.92 respectively.

COMPARISON WITH EXISTING METHODS

The DSVM framework outperforms classification with more standard multi-class "one-against-all" SVM and linear-discriminant analysis.

CONCLUSION

The promising results of the proposed methodology suggest that it may be a valuable alternative to existing automatic methods and that it could accelerate visual scoring by providing a robust starting hypnogram that can be further fine-tuned by expert inspection.

Amplitude normalization applied to an artificial neural network-based automatic sleep spindle detection system

Sleep spindles are significant rhythmic transients present in the sleep electroencephalogram (EEG) of non-rapid eye movement (NREM) sleep. Automatic sleep spindle detection techniques are sought for the automation of sleep staging and the detailed study of sleep spindle patterns, of possible physiological significance. A deficiency of many of the available automatic detection techniques is their reliance on the amplitude level of the recorded EEG voltage values. In the present work, an automatic sleep spindle detection system that has been previously proposed, using a Multi-Layer Perceptron (MLP) Artificial Neural Network (ANN), was evaluated using a voltage amplitude normalization procedure, with the aim of making the performance of the ANN independent of the absolute voltage level of the individual subjects' recordings. The application of the normalization procedure led to a reduction in the false positive rate (FPR) as well as in the sensitivity. When the ANN was trained on a combination of data from healthy subjects, the reduction of FPR was from 42.6% to 19%, while the sensitivity of the ANN was kept at acceptable levels, i.e., 73.4% for the normalized procedure vs 84.6% for the non-normalized procedure.

Evaluating the use of line length for automatic sleep spindle detection

Sleep spindles are transient waveforms observed on the electroencephalogram (EEG) during the N2 stage of sleep. In this paper we evaluate the use of line length, an efficient and low-complexity time domain feature, for automatic detection of sleep spindles. We use this feature with a simple algorithm to detect spindles achieving sensitivity of 83.6% and specificity of 87.9%. We also present a comparison of these results with other spindle detection methods evaluated on the same dataset. Further, we implemented the algorithm on a MSP430 microcontroller achieving a power consumption of 56.7 μW. The overall detection performance, combined with the low power consumption show that line length could be a useful feature for detecting sleep spindles in wearable and resource-constrained systems.