EEG Samples Research Articles

A convolutional neural network for sleep stage scoring from raw single-channel EEG

Abstract We present a novel method for automatic sleep scoring based on single-channel EEG. We introduce the use of a deep convolutional neural network (CNN) on raw EEG samples for supervised learning of 5-class sleep stage prediction. The network has 14 layers, takes as input the 30-s epoch to be classified as well as two preceding epochs and one following epoch for temporal context, and requires no signal preprocessing or feature extraction phase. We train and evaluate our system using data from the Sleep Heart Health Study (SHHS), a large multi-center cohort study including expert-rated polysomnographic records. Performance metrics reach the state of the art, with accuracy of 0.87 and Cohen kappa of 0.81. The use of a large cohort with multiple expert raters guarantees good generalization. Finally, we present a method for visualizing class-wise patterns learned by the network.

The validation of simplified EEG derived from the bispectral index monitor in post-cardiac arrest patients

AimsWe aimed to validate retrospectively the accuracy of simplified electroencephalography (EEG) monitoring derived from the bispectral index (BIS) monitor in post-cardiac arrest (CA) patients. MethodsSuccessfully resuscitated CA patients were transferred to the Catherization Lab followed by percutaneous coronary intervention when indicated. On arrival at the coronary care unit, bilateral BIS monitoring was started and continued up to 72 h. Raw simplified EEG tracings were extracted from the BIS monitor at a time point coinciding with the registration of standard EEG monitoring. BIS EEG tracings were reviewed by two neurophysiologists, who were asked to indicate the presence of following patterns: diffuse slowing rhythm, burst suppression pattern, cerebral inactivity, periodic epileptiform discharges and status epilepticus (SE). Additionally, these simplified BIS EEG tracings were analysed by two inexperienced investigators, who were asked to indicate the presence of SE only. ResultsThirty-two simplified BIS EEG samples were analysed. Compared to standard EEG, neurophysiologists interpreted all simplified EEG samples with a sensitivity of 86%, a specificity of 100% and an interobserver variability of 0.843. Furthermore, SE was identified with a sensitivity of 80% and a specificity of 94% by two unexperienced physicians. ConclusionUsing a simple classification system, raw simplified EEG derived from a BIS monitoring device is comparable to standard EEG monitoring. Moreover, investigators without EEG experience were capable to identify SE in post-CA patients. Future studies will be warranted to confirm our results and to determine the added value of using simplified BIS EEG in terms of prognostic and therapeutic implications.

Effects of halothane on the electroencephalogram of the chicken.

Little is known about the effects of inhalant anaesthetics on the avian electroencephalogram (EEG). The effects of halothane on the avian EEG are of interest, as this agent has been widely used to study nociception and analgesia in mammals. The objective of this study was to characterize the effects of halothane anaesthesia on the EEG of the chicken. Twelve female Hyline Brown chickens aged 8–10 weeks were anaesthetized with halothane in oxygen. For each bird, anaesthesia was progressively increased from 1–1.5 to 2 times the Minimum Anesthetic Concentration (MAC), then progressively decreased again. At each concentration, a sample of EEG was recorded after a 10‐min stabilization period. The mean Total Power (PTOT), Median Frequency (F50) and 95% Spectral Edge Frequency (F95) were calculated at each halothane MAC, along with the Burst Suppression Ratio (BSR). Burst suppression was rare and BSR did not differ between halothane concentrations. Increasing halothane concentration from 1 to 2 MAC resulted in a decrease in F50 and increase in PTOT, while F95 increased when MAC was reduced from 1.5 to 1. The results indicate dose‐dependent spectral EEG changes consistent with deepening anaesthesia in response to increasing halothane MAC. As burst suppression was rare, even at 1.5 or 2 times MAC, halothane may be a suitable anaesthetic agent for use in future studies exploring EEG activity in anaesthetized birds.

Diagnostic utility of eight-channel EEG for detecting generalized or hemispheric seizures and rhythmic periodic patterns

ObjectivesTo compare the diagnostic utility of electroencephalography (EEG) using reduced, 8-channel montage (rm-EEG) to full, 18-channel montage (fm-EEG) for detection of generalized or hemispheric seizures and rhythmic periodic patterns (RPPs) by neurologists with extensive EEG training, neurology residents with minimal EEG exposure, and medical students without EEG experience. MethodsWe presented EEG samples in both fm-EEG (bipolar montage) and rm-EEG (lateral leads of bipolar montage) to 20 neurologists, 20 residents, and 42 medical students. Unanimous agreement of three senior epileptologists defined samples as seizures (n = 7), RPPs (n = 10), and normal or slowing (n = 20). Differences in median accuracy, sensitivity, and specificity were assessed using Wilcoxon signed-rank tests. ResultsFull and reduced EEG demonstrated similar accuracy when read by neurologists (fm-EEG: 95%, rm-EEG: 95%, p = 0.29), residents (fm-EEG: 80%, rm-EEG: 80%, p = 0.05), and students (fm-EEG: 60%, rm-EEG: 51%, p = 0.68). Moreover, neurologists’ sensitivity for detecting seizure activity was comparable between fm-EEG (100%) and rm-EEG (98%) (p = 0.17). Furthermore, the specificity of rm-EEG for seizures and RPP (neurologists: 100%, residents: 90%, students: 86%) was significantly greater than that of fm-EEG (neurologists: 93%, p = 0.03; residents: 80%, p = 0.01; students: 69%, p < 0.001). ConclusionsThe reduction of the number of EEG channels from 18 to 8 does not compromise neurologists’ sensitivity for detecting seizures that are often a core reason for performing urgent EEG. It may also increase their specificity for detecting rhythmic and periodic patterns, and thereby providing important diagnostic information to guide patient’s management. SignificanceOur study is the first to document the utility of a reduced channel EEG above the hairline compared to full montage EEG in aiding medical staff with varying degrees of EEG training to detect generalized or hemispheric seizures.

Safe Semi-Supervised Extreme Learning Machine for EEG Signal Classification

One major challenge in the current brain–computer interface research is the accurate classification of time-varying electroencephalographic (EEG) signals. The labeled EEG samples are usually scarce, while the unlabeled samples are available in large quantities and easy to collect in real applications. Semi-supervised learning (SSL) methods can utilize both labeled and unlabeled data to improve performance over supervised approaches. However, it has been reported that the unlabeled data may undermine the performance of SSL in some cases. To improve the safety of SSL, we proposed a new safety-control mechanism by analyzing the differences between unlabeled data analysis in supervised and semi-supervised learning. We then develop and implement a safe classification method based on the semi-supervised extreme learning machine (SS-ELM). Following this approach, the Wasserstein distance is used to measure the similarities between the predictions obtained from ELM and SS-ELM algorithms, and a different risk degree is thereby calculated for each unlabeled data instance. A risk-based regularization term is then constructed and embedded into the objective function of the SS-ELM. Extensive experiments were conducted using benchmark and EEG datasets to evaluate the effectiveness of the proposed method. Experimental results show that the performance of the new algorithm is comparable to SS-ELM and superior to ELM on average. It is thereby shown that the proposed method is safe and efficient for the classification of EEG signals.

Investigation of True High Frequency Electrical Substrates of fMRI-Based Resting State Networks Using Parallel Independent Component Analysis of Simultaneous EEG/fMRI Data.

Previous work using simultaneously acquired electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) data has shown that the slow temporal dynamics of resting state brain networks (RSNs), e.g., default mode network (DMN), visual network (VN), obtained from fMRI are correlated with smoothed and down sampled versions of various EEG features such as microstates and band-limited power envelopes. Therefore, even though the down sampled and smoothed envelope of EEG gamma band power is correlated with fMRI fluctuations in the RSNs, it does not mean that the electrical substrates of the RSNs fluctuate with periods <100 ms. Based on the scale free properties of EEG microstates and their correlation with resting state fMRI fluctuations in the RSNs, researchers have speculated that truly high frequency electrical substrates may exist for the RSNs, which would make resting fluctuations obtained from fMRI more meaningful to typically occurring fast neuronal processes in the sub-100 ms time scale. In this study, we test this critical hypothesis using an integrated framework involving simultaneous EEG/fMRI acquisition, fast fMRI sampling (TR = 200 ms) using multiband EPI (MB EPI), and EEG/fMRI fusion using parallel independent component analysis (pICA) which does not require the down sampling of EEG to fMRI temporal resolution. Our results demonstrate that with faster sampling, high frequency electrical substrates (fluctuating with periods <100 ms time scale) of the RSNs can be observed. This provides a sounder neurophysiological basis for the RSNs.

The effect of increased intracranial EEG sampling rates in clinical practice

ObjectiveRecent research suggests that high frequency intracranial EEG (iEEG) may improve localization of epileptic networks. This study aims to determine whether recording macroelectrode iEEG with higher sampling rates improves seizure localization in clinical practice. Methods14 iEEG seizures from 10 patients recorded with >2000 Hz sampling rate were downsampled to four sampling rates: 100, 200, 500, 1000 Hz. In the 56 seizures, seizure onset time and location was marked by 5 independent, blinded EEG experts. ResultsWhen reading iEEG under clinical conditions, there was no consistent difference in time or localization of seizure onset or number of electrodes involved in the seizure onset zone with sampling rates varying from 100 to 1000 Hz. Stratification of patients by outcome did not improve with higher sampling rate. ConclusionWhen utilizing standard clinical protocols, there was no benefit to acquiring iEEGs with sampling rate >100 Hz. Significant variability was noted in EEG marking both within and between individual expert EEG readers. SignificanceAlthough commercial equipment is capable of sampling much faster than 100 Hz, tools allowing visualization of subtle high frequency activity such as HFOs will be required to improve patient care. Quantitative methods may decrease reader variability, and potentially improve patient outcomes.

The use of single bipolar scalp derivation for the detection of ictal events during long-term EEG monitoring.

Epilepsy is difficult to diagnose using routine EEG recordings of short duration in patients who have low seizure frequency. Long-term EEG may be useful but is impractical in an out-of-hospital setting. We investigated whether single-channel scalp EEG placed behind the earlobe is suitable for seizure identification during prolonged EEG monitoring. Scalp EEG samples were selected from subjects over 15 years of age, and comprised two segments of either background followed by seizure or background followed by background. Bipolar EEG derivations in three directions (F8-T8, C4-T8 and T8-P8) were evaluated for the presence of a seizure by two experienced reviewers. For each EEG segment containing a seizure, one pair of electrodes was oriented towards the suspected region of seizure onset, while two pairs of electrodes were oriented elsewhere. The EEG data contained five frontally localized seizures, five parietal, five temporal, two occipital, and four primary or secondary generalized seizures. The sensitivity and specificity for recognition of seizures was 86% and 95% for Reviewer 1, and 79% and 99% for Reviewer 2, respectively. When identifying a seizure with the lead orientation towards the region of seizure onset, both reviewers identified 20 out of 21 seizures (95%). When the lead was not oriented towards the region of seizure onset, the reviewers identified 34 and 30 out of 42 ictal records correctly, respectively. These results suggest that it is possible to identify epileptic seizures by bipolar EEG derivation using only two scalp electrodes. Lead orientation towards the suspected region of seizure onset is important for optimal detection sensitivity.

Unbiased estimation of permutation entropy in EEG analysis for Alzheimer's disease classification

Abstract The EEG signal of healthy patient can be recognized as an output of a chaotic system. There are many measures of chaotic behaviour: Hurst and Lyapunov exponents, various dimensions of attractor, various entropy measures, etc. We prefer permutation entropy of equidistantly sampled data. The novelty of our approach is in bias reduction of permutation entropy estimates, memory decrease, and time complexities of permutation analysis. Therefore, we are not limited by the EEG signal and permutation sample lengths. This general method was used for channel by channel analysis of Alzheimer's diseased (AD) and healthy (CN) patients to point out the differences between AD and CN groups. Our technique also enables to study the influence of EEG sampling frequency in a wide range. The best results were obtained for sampling frequency 200 Hz, using at most window of length 10. In the case of Alzheimer's disease, we observed a statistically significant decrease in permutation entropy at all channels.

O23 Characteristic slow gamma and delta frequency changes in the preictal EEG of partial epileptic patients

The preictal EEG has informations about the onset of the epileptic seizure. Authors aim was to find the characteristic EEG frequency changes which predict the launch of the seizure. Authors analysed the slow delta and gamma oscillations (0–100 Hz ) in the preictal scalp EEG of partial epileptic patients. The lower delta frequency was measured 0–2.5 Hz, and the lower gamma frequency was measured 35–100 Hz. It were applied the Main Dominant Frequency, the Median Frequency and the Relative Power frequency analysis methods. The Relative Power frequency analysis proportionated the results of the spectrum. They have been selected 12 electroclinical seizures from the presurgical video EEG monitorization of 5 patients The measurement was done from the 3 min long preictal EEG sample. There were not analysed the seizure EEG. Authors observed in the 3 min long examined EEG period, that for several seconds the frequency values of the electrodes time to time approach each other. This phenomenon was much more pronounced on the electrodes of the ictal hemisphere. There were specific moments (1 or 2 s) of the preictal EEG length, when the frequency values of the electrodes were almost the same. In the point of view of the brain network theory our results are the parameters of the synchronisation in the epileptic network. There are prognostical parameters in the frequency spectrum of the preictal 3 min long EEG period. The authors have been found characteristic frequency changes in the preictal EEG.

A novel HFO-based method for unsupervised localization of the seizure onset zone in drug-resistant epilepsy.

High frequency oscillations (HFOs) are potential biomarkers of epileptic areas. In patients with drug-resistant epilepsy, HFO rates tend to be higher in the seizure onset zone (SOZ) than in other brain regions and the resection of HFO-generating areas positively correlates with seizure-free surgery outcome. Nonetheless, the development of robust unsupervised HFO-based tools for SOZ localization remains challenging. Current approaches predict the SOZ by processing small samples of intracranial EEG (iEEG) data and applying patient-specific thresholds on the HFO rate. The HFO rate, though, varies largely over time with the patient's conditions (e.g., sleep versus wakefulness) and across patients. We propose a novel localization method for SOZ that uses a time-varying, HFO-based index to estimate the epileptic susceptibility of the iEEG channels. The method is insensitive to the average HFO rate across channels (which is both patient- and condition-specific), tracks the channel susceptibility over time, and predicts the SOZ based on the temporal evolution of the HFO rate. Tested on a preliminary dataset of continuous multi-day multichannel interictal iEEG recordings from two epileptic patients (117±97.6 h/per patient, mean ± S.D.), the reported SOZ prediction had an average 0.70±0.18 accuracy and 0.67±0.07 area under the ROC curve (mean ± S.D.) across patients.

A flexible method for the automated offline-detection of artifacts in multi-channel electroencephalogram recordings.

Electroencephalogram (EEG) signal quality is often compromised by artifacts that corrupt quantitative EEG measurements used in clinical applications and EEG-related studies. Techniques such as filtering, regression analysis and blind source separation are often used to remove these artifacts. However, these preprocessing steps do not allow for complete artifact correction. We propose a method for the automated offline-detection of remaining artifacts after preprocessing in multi-channel EEG recordings. In contrast to existing methods it requires neither adaptive parameters varying between recordings nor a topography template. It is suited for short EEG segments and is flexible with regard to target applications. The algorithm was developed and tested on 60 clinical EEG samples of 20 seconds each that were recorded both in resting state and during cognitive activation to gain a realistic artifact set. Five EEG features were used to quantify temporal and spatial signal variations. Two distance measures for the single-channel and multi-channel variations of these features were defined. The global thresholds were determined by three-fold cross-validation and Youden's J statistic in conjunction with receiver operating characteristics (ROC curves). We observed high sensitivity of 95.5%±4.8 and specificity of 88.8%±2.1. The method has thus shown great potential and is promising as a possible tool for both EEG-based clinical applications and EEG-related research.

실용적 뇌파 기반 사용자 인증을 위한 단일 채널 EEG 측정 장비를 통해 수집된 EEG 샘플의 점진적 제거 방법

실용적 뇌파 기반 사용자 인증을 위한 단일 채널 EEG 측정 장비를 통해 수집된 EEG 샘플의 점진적 제거 방법

Decrease of global current source density predicts successful treatment in absence and juvenile myoclonic epilepsies

ObjectiveTo investigate relationship between treatment efficiency and EEG background activity changes in absence epilepsy (AE) and juvenile myoclonic epilepsy (JME) patients. Patients and methodsEEGs of 31 patients were analysed before treatment and after six months of treatment. Three minutes of artifact-free waking EEG background activity (without epileptiform potentials) were analysed for each patient in both conditions. All the EEG samples were processed to LORETA (Low Resolution Electromagnetic Tomography). Average of all the voxel-wise current source density (CSD) values within the 0.5–8.0Hz frequency range was computed for each EEG. Fischer's exact test was used to investigate association between the global CSD changes and the therapeutic outcome. ResultsTight connection was demonstrated between seizure freedom and decreased CSD, and between persisting seizures and increased CSD (p<0.001). SignificanceAn EEG-based biomarker that predicts successful drug treatment was described.

Features of EEG Activity Related to Realization of Cyclic Unimanual and Bimanual Hand Movements in Humans

In 10 tested dextral subjects, EEG activity was recorded during the performance of unimanual and bimanual cyclic movements of the hands and fingers. The movements corresponded to clenching the fingers into a fist and the subsequent unclenching of the fingers. The test consisted of four successive stages, the resting state, movement of the left hand, that of the right hand, and movement of both hands. The dependences between the spectral power and coherence of the respective EEG samples on the type of the test performed, on the type of the movement (uni- or bimanual), and on the laterality of the latter in the case of the unimanual movement were examined. The results obtained allow us to propose the following conclusions: (i) α and β EEG rhythms are characterized by different functional importance with respect to manual motor activity; (ii) neural control of bimanual movements cannot be considered “a sum of the controls” of unimanual movements, and (iii) control of bimanual movements may be largely based on the control of the movement by a subdominant upper limb.

Variability of ICA decomposition may impact EEG signals when used to remove eyeblink artifacts.

Despite the growing use of independent component analysis (ICA) algorithms for isolating and removing eyeblink-related activity from EEG data, we have limited understanding of how variability associated with ICA uncertainty may be influencing the reconstructed EEG signal after removing the eyeblink artifact components. To characterize the magnitude of this ICA uncertainty and to understand the extent to which it may influence findings within ERP and EEG investigations, ICA decompositions of EEG data from 32 college-aged young adults were repeated 30 times for three popular ICA algorithms. Following each decomposition, eyeblink components were identified and removed. The remaining components were back-projected, and the resulting clean EEG data were further used to analyze ERPs. Findings revealed that ICA uncertainty results in variation in P3 amplitude as well as variation across all EEG sampling points, but differs across ICA algorithms as a function of the spatial location of the EEG channel. This investigation highlights the potential of ICA uncertainty to introduce additional sources of variance when the data are back-projected without artifact components. Careful selection of ICA algorithms and parameters can reduce the extent to which ICA uncertainty may introduce an additional source of variance within ERP/EEG studies.

Mental State Recognition by using Brain Waves

Mental State classification plays vital role in communication between human and computer. It can be classified by brain waves. Brain signals recording methods along scalp are invasive and non-invasive. The advanced common method is noninvasive. The 10-10 and 10-20 electrodes system standards are used in the non-invasive method. It is aimed to find Mental State of a sample by using EEG signals. In this work EEG sample are collected from the DEAP database. 30 preprocessed EEG signals from DEAP data sets are worked to get different frequency band by Welch method. Power spectral density in alpha and beta band is tabulated. The valence arousal points from the pre-processed data is obtained from the matlab files Four mental states are classified as four quadrants of graph. 30 participants are classified into four mental states namely happy, stressed, depressed and relaxed in MATLAB. As result four mental states are classified as four quadrants of graph.

EEG classification for motor imagery and resting state in BCI applications using multi-class Adaboost extreme learning machine.

Brain-computer interface (BCI) systems provide an alternative communication and control approach for people with limited motor function. Therefore, the feature extraction and classification approach should differentiate the relative unusual state of motion intention from a common resting state. In this paper, we sought a novel approach for multi-class classification in BCI applications. We collected electroencephalographic (EEG) signals registered by electrodes placed over the scalp during left hand motor imagery, right hand motor imagery, and resting state for ten healthy human subjects. We proposed using the Kolmogorov complexity (Kc) for feature extraction and a multi-class Adaboost classifier with extreme learning machine as base classifier for classification, in order to classify the three-class EEG samples. An average classification accuracy of 79.5% was obtained for ten subjects, which greatly outperformed commonly used approaches. Thus, it is concluded that the proposed method could improve the performance for classification of motor imagery tasks for multi-class samples. It could be applied in further studies to generate the control commands to initiate the movement of a robotic exoskeleton or orthosis, which finally facilitates the rehabilitation of disabled people.

Classification of epilepsy seizure phase using interval type-2 fuzzy support vector machines

An interval type-2 fuzzy support vector machine (IT2FSVM) is proposed to solve a classification problem which aims to classify three epileptic seizure phases (seizure-free, pre-seizure and seizure) from the electroencephalogram (EEG) captured from patients with neurological disorder symptoms. The effectiveness of the IT2FSVM classifier is evaluated based on a set of EEG samples which are collected from 10 patients at Peking university hospital. The EEG samples for the three seizure phases were captured by the 112 2-s 19 channel EEG epochs, where each patient was extracted for each sample. Feature extraction was used to reduce the feature vector of the EEG samples to 45 elements and the EEG samples with the reduced features are used for training the IT2FSVM classifier. The classification results obtained by the IT2FSVM are compared with three traditional classifiers namely Support Vector Machine, k-Nearest Neighbor and naive Bayes. The experimental results show that the IT2FSVM classifier is able to achieve superior learning capabilities with respect to the uncontaminated samples when compared with the three classifiers. In order to validate the level of robustness of the IT2FSVM, the original EEG samples are contaminated with Gaussian white noise at levels of 0.05, 0.1, 0.2 and 0.5. The simulation results show that the IT2FSVM classifier outperforms the traditional classifiers under the original dataset and also shows a high level of robustness when compared to the traditional classifiers with white Gaussian noise applied to it.

Convolutional Neural Network for Multi-Category Rapid Serial Visual Presentation BCI.

Brain computer interfaces rely on machine learning (ML) algorithms to decode the brain's electrical activity into decisions. For example, in rapid serial visual presentation (RSVP) tasks, the subject is presented with a continuous stream of images containing rare target images among standard images, while the algorithm has to detect brain activity associated with target images. Here, we continue our previous work, presenting a deep neural network model for the use of single trial EEG classification in RSVP tasks. Deep neural networks have shown state of the art performance in computer vision and speech recognition and thus have great promise for other learning tasks, like classification of EEG samples. In our model, we introduce a novel spatio-temporal regularization for EEG data to reduce overfitting. We show improved classification performance compared to our earlier work on a five categories RSVP experiment. In addition, we compare performance on data from different sessions and validate the model on a public benchmark data set of a P300 speller task. Finally, we discuss the advantages of using neural network models compared to manually designing feature extraction algorithms.