Automated real-time epileptic seizure detection in scalp EEG recordings using an algorithm based on wavelet packet transform

Dynamic Neural State Identification in Deep Brain Local Field Potentials of Neuropathic Pain

In neuropathic pain, the neurophysiological and neuropathological function of the ventro-posterolateral nucleus of the thalamus (VPL) and the periventricular gray/periaqueductal gray area (PVAG) involves multiple frequency oscillations. Moreover, oscillations related to pain perception and modulation change dynamically over time. Fluctuations in these neural oscillations reflect the dynamic neural states of the nucleus. In this study, an approach to classifying the synchronization level was developed to dynamically identify the neural states. An oscillation extraction model based on windowed wavelet packet transform was designed to characterize the activity level of oscillations. The wavelet packet coefficients sparsely represented the activity level of theta and alpha oscillations in local field potentials (LFPs). Then, a state discrimination model was designed to calculate an adaptive threshold to determine the activity level of oscillations. Finally, the neural state was represented by the activity levels of both theta and alpha oscillations. The relationship between neural states and pain relief was further evaluated. The performance of the state identification approach achieved sensitivity and specificity beyond 80% in simulation signals. Neural states of the PVAG and VPL were dynamically identified from LFPs of neuropathic pain patients. The occurrence of neural states based on theta and alpha oscillations were correlated to the degree of pain relief by deep brain stimulation. In the PVAG LFPs, the occurrence of the state with high activity levels of theta oscillations independent of alpha and the state with low-level alpha and high-level theta oscillations were significantly correlated with pain relief by deep brain stimulation. This study provides a reliable approach to identifying the dynamic neural states in LFPs with a low signal-to-noise ratio by using sparse representation based on wavelet packet transform. Furthermore, it may advance closed-loop deep brain stimulation based on neural states integrating multiple neural oscillations.

A Novel Approach for Real-Time Recognition of Epileptic Seizures Using Minimum Variance Modified Fuzzy Entropy

OBJECTIVE

Validation of epileptic seizures annotations from long-term electroencephalogram (EEG) recordings is a tough and tedious task for the neurological community. It is a well-known fact that computerized qualitative methods thoroughly assess the complex brain dynamics toward seizure detection and proven as one of the acceptable clinical indicators.

METHODS

This research study suggests a novel approach for real-time recognition of epileptic seizure from EEG recordings by a technique referred as minimum variance modified fuzzy entropy (MVMFzEn). Multichannel EEG recordings of 4.36 h of epileptic seizures and 25.74 h of normal EEG were considered. Signal processing techniques such as filters and independent component analysis were appropriated to reduce noise and artifacts. Unlike, the predefined fuzzy membership function, the modified fuzzy entropy utilizes relative energy as a membership function followed by scaling operation to obtain the feature.

RESULTS

Results revealed that MVMFzEn drops abruptly during an epileptic activity and this fact was used to set a threshold. An automated threshold derived from MVMFzEn assesses the classification efficiency of the given data during validation. It was observed from the results that the proposed method yields a classification accuracy of 100% without the use of any classifier.

CONCLUSION

The graphical user interface was designed in MATLAB to automatically label the normal and epileptic segments in the long-term EEG recordings.

SIGNIFICANCE

The ground truth clinical validation using validation specificity and validation sensitivity confirms the suitability of the proposed technique for automated annotation of epileptic seizures in real time.

Wavelet-based feature extraction for classification of epileptic seizure EEG signal

Electroencephalogram (EEG) signal-processing techniques are the prominent role in the detection and prediction of epileptic seizures. The detection of epileptic activity is cumbersome and needs a detailed analysis of the EEG data. Therefore, an efficient method for classifying EEG data is required. In this work, a constructive pattern recognition strategy for analysing EEG data as normal and epileptic seizure has been proposed. With this strategy, the signals were decomposed into frequency sub-bands using discrete wavelet transform (DWT). principal component analysis (PCA) and linear discriminant analysis (LDA) are applied to reduce the dimensionality of EEG data. These reduced features were used as input to Naïve Bayes and K-Nearest Neighbour Classifier to classify normal or epileptic seizure signal. The performance of classifier was evaluated in terms of accuracy, sensitivity and specificity. The experimental results show that PCA with Naïve Bayes classifier provides 98.6% accuracy and LDA with Naïve Bayes classifier attains improved result of 99.8% accuracy. Also, the result shows that PCA, LDA with K-NN achieves 98.5% and 100% accuracy. This evaluation is used to propose a reliable, practical epilepsy detection method to enhance the patient's care and quality of life.

Kernel-Based Relevance Analysis with Enhanced Interpretability for Detection of Brain Activity Patterns

We introduce Enhanced Kernel-based Relevance Analysis (EKRA) that aims to support the automatic identification of brain activity patterns using electroencephalographic recordings. EKRA is a data-driven strategy that incorporates two kernel functions to take advantage of the available joint information, associating neural responses to a given stimulus condition. Regarding this, a Centered Kernel Alignment functional is adjusted to learning the linear projection that best discriminates the input feature set, optimizing the required free parameters automatically. Our approach is carried out in two scenarios: (i) feature selection by computing a relevance vector from extracted neural features to facilitating the physiological interpretation of a given brain activity task, and (ii) enhanced feature selection to perform an additional transformation of relevant features aiming to improve the overall identification accuracy. Accordingly, we provide an alternative feature relevance analysis strategy that allows improving the system performance while favoring the data interpretability. For the validation purpose, EKRA is tested in two well-known tasks of brain activity: motor imagery discrimination and epileptic seizure detection. The obtained results show that the EKRA approach estimates a relevant representation space extracted from the provided supervised information, emphasizing the salient input features. As a result, our proposal outperforms the state-of-the-art methods regarding brain activity discrimination accuracy with the benefit of enhanced physiological interpretation about the task at hand.

Spike densities of the amygdala and neocortex reflect progression of kindled motor seizures

Amygdala kindling is a common temporal lobe-like seizure model. In the present study, temporal and spectral analyses of the ictal period were investigated throughout amygdala kindling in response to different behavioral seizures. Right-side amygdala was kindled to induce epileptiform afterdischarges (ADs). ADs of both the frontal cortex and amygdala were analyzed. Powers of the low (0-9 Hz)- and high (12-30 Hz)-frequency bands in response to different behavioral seizures were calculated. Densities of upward and downward peaks of spikes, which reflected information of spike count and spike pattern, throughout kindle-induced ADs were calculated. Progression was seen in the temporal and spectral characteristics of amygdala-kindled ADs in response to behaviors. Numbers of significant differences of all 1-s AD segments between two Racine's seizure stages were significantly higher in upward and downward indexes of the temporal spike than those using the spectral method in both the amygdala and neocortex. Ability for distinguishing seizure stages was significantly higher in temporal spike density of amygdala ADs compared to those of frontal ADs. Our results showed that amygdala kindling caused spectrotemporal changes of activities in the amygdala and frontal cortex. The density of spike-related peaks had better distinguishability in response to behavioral seizures, particularly in a seizure zone of amygdala. The present study provides a new temporal index of spike's peak density to understand progression of motor seizures in the kindling process.

An extensive review on development of EEG-based computer-aided diagnosis systems for epilepsy detection

Epilepsy is considered as fourth most prominent neurological disorder in the world that can affect people of all age groups. Currently, around 65 million people throughout the world are suffering from epilepsy. It is evident that electroencephalograph (EEG) signals are most commonly used for detection of epileptic seizures but today many modern techniques have been developed to analyze underlying features of these EEG signals. As EEG contains a large amount of complicated information, so many researchers are trying to develop automatic systems for complete feature extraction. This paper provides a generalized review and performance comparison of popular seizure detection algorithms that are developed in the last decade. The main objective of this paper is to briefly discuss all existing developments in the field of computer-aided diagnosis system for epilepsy detection so that future researchers can find a better track for the new invention.

A high-performance seizure detection algorithm based on Discrete Wavelet Transform (DWT) and EEG

In the past decade, Discrete Wavelet Transform (DWT), a powerful time-frequency tool, has been widely used in computer-aided signal analysis of epileptic electroencephalography (EEG), such as the detection of seizures. One of the important hurdles in the applications of DWT is the settings of DWT, which are chosen empirically or arbitrarily in previous works. The objective of this study aimed to develop a framework for automatically searching the optimal DWT settings to improve accuracy and to reduce computational cost of seizure detection. To address this, we developed a method to decompose EEG data into 7 commonly used wavelet families, to the maximum theoretical level of each mother wavelet. Wavelets and decomposition levels providing the highest accuracy in each wavelet family were then searched in an exhaustive selection of frequency bands, which showed optimal accuracy and low computational cost. The selection of frequency bands and features removed approximately 40% of redundancies. The developed algorithm achieved promising performance on two well-tested EEG datasets (accuracy >90% for both datasets). The experimental results of the developed method have demonstrated that the settings of DWT affect its performance on seizure detection substantially. Compared with existing seizure detection methods based on wavelet, the new approach is more accurate and transferable among datasets.

Principal dynamic mode analysis of neural mass model for the identification of epileptic states

The detection of epileptic seizures in Electroencephalography (EEG) signals is significant for the diagnosis and treatment of epilepsy. In this paper, in order to obtain characteristics of various epileptiform EEGs that may differentiate different states of epilepsy, the concept of Principal Dynamic Modes (PDMs) was incorporated to an autoregressive model framework. First, the neural mass model was used to simulate the required intracerebral EEG signals of various epileptiform activities. Then, the PDMs estimated from the nonlinear autoregressive Volterra models, as well as the corresponding Associated Nonlinear Functions (ANFs), were used for the modeling of epileptic EEGs. The efficient PDM modeling approach provided physiological interpretation of the system. Results revealed that the ANFs of the 1st and 2nd PDMs for the auto-regressive input exhibited evident differences among different states of epilepsy, where the ANFs of the sustained spikes' activity encountered at seizure onset or during a seizure were the most differentiable from that of the normal state. Therefore, the ANFs may be characteristics for the classification of normal and seizure states in the clinical detection of seizures and thus provide assistance for the diagnosis of epilepsy.

Circadian variation of scalp EEG: a novel measure based on wavelet packet transform and differential entropy

We propose a novel entropy-based measure to quantify the circadian variations of scalp electroencephalogram (EEG) by analyzing waking epochs of nap opportunities under an ultradian sleep-wake cycle (USW) protocol. To compute this circadian measure for a nap opportunity, each waking epoch (~1 sec) is decomposed using wavelet packet transform and the relative energy for the desired frequency band (here, 10-12 Hz) is calculated. Then, in a bootstrapping procedure, a shape statistic (skewness or kurtosis) of the relative energy distribution, after each resampling, is computed. Finally, the probability density function of this statistic is estimated, and the corresponding differential entropy is considered as the circadian measure. This measure was evaluated using EEG recordings from 4 healthy subjects during a 72-h USW procedure. According to the results, the proposed measure showed a significant circadian variation both for individual and group data, with peak values occurring near the core body temperature minimum. The performance of the entropy-based measure was also compared with that of two other measures, namely mean energy logarithm and mean energy ratio, revealing the superiority of this measure.

Identification of scalp EEG circadian variation using a novel correlation sum measure

OBJECTIVE

In this paper, we propose a novel method to determine the circadian variation of scalp electroencephalogram (EEG) in both individual and group levels using a correlation sum measure, quantifying self-similarity of the EEG relative energy across waking epochs.

APPROACH

We analysed EEG recordings from central-parietal and occipito-parietal montages in nine healthy subjects undergoing a 72 h ultradian sleep-wake cycle protocol. Each waking epoch (∼ 1 s) of every nap opportunity was decomposed using the wavelet packet transform, and the relative energy for that epoch was calculated in the desired frequency band using the corresponding wavelet coefficients. Then, the resulting set of energy values was resampled randomly to generate different subsets with equal number of elements. The correlation sum of each subset was then calculated over a range of distance thresholds, and the average over all subsets was computed. This average value was finally scaled for each nap opportunity and considered as a new circadian measure.

MAIN RESULTS

According to the evaluation results, a clear circadian rhythm was identified in some EEG frequency ranges, particularly in 4-8 Hz and 10-12 Hz. The correlation sum measure not only was able to disclose the circadian rhythm on the group data but also revealed significant circadian variations in most individual cases, as opposed to previous studies only reporting the circadian rhythms on a population of subjects. Compared to a naive measure based on the EEG absolute energy in the frequency band of interest, the proposed measure showed a clear superiority using both individual and group data. Results also suggested that the acrophase (i.e., the peak) of the circadian rhythm in 10-12 Hz occurs close to the core body temperature minimum.

SIGNIFICANCE

These results confirm the potential usefulness of the proposed EEG-based measure as a non-invasive circadian marker.

Online Epileptic Seizure Prediction Using Wavelet-Based Bi-Phase Correlation of Electrical Signals Tomography

Considerable efforts have been made in order to predict seizures. Among these methods, the ones that quantify synchronization between brain areas, are the most important methods. However, to date, a practically acceptable result has not been reported. In this paper, we use a synchronization measurement method that is derived according to the ability of bi-spectrum in determining the nonlinear properties of a system. In this method, first, temporal variation of the bi-spectrum of different channels of electro cardiography (ECoG) signals are obtained via an extended wavelet-based time-frequency analysis method; then, to compare different channels, the bi-phase correlation measure is introduced. Since, in this way, the temporal variation of the amount of nonlinear coupling between brain regions, which have not been considered yet, are taken into account, results are more reliable than the conventional phase-synchronization measures. It is shown that, for 21 patients of FSPEEG database, bi-phase correlation can discriminate the pre-ictal and ictal states, with very low false positive rates (FPRs) (average: 0.078/h) and high sensitivity (100%). However, the proposed seizure predictor still cannot significantly overcome the random predictor for all patients.

Identification and monitoring of brain activity based on stochastic relevance analysis of short-time EEG rhythms

Background

The extraction of physiological rhythms from electroencephalography (EEG) data and their automated analyses are extensively studied in clinical monitoring, to find traces of interictal/ictal states of epilepsy.

Methods

Because brain wave rhythms in normal and interictal/ictal events, differently influence neuronal activity, our proposed methodology measures the contribution of each rhythm. These contributions are measured in terms of their stochastic variability and are extracted from a Short Time Fourier Transform to highlight the non–stationary behavior of the EEG data. Then, we performed a variability–based relevance analysis by handling the multivariate short–time rhythm representation within a subspace framework. This maximizes the usability of the input information and preserves only the data that contribute to the brain activity classification. For neural activity monitoring, we also developed a new relevance rhythm diagram that qualitatively evaluates the rhythm variability throughout long time periods in order to distinguish events with different neuronal activities.

Results

Evaluations were carried out over two EEG datasets, one of which was recorded in a noise–filled environment. The method was evaluated for three different classification problems, each of which addressed a different interpretation of a medical problem. We perform a blinded study of 40 patients using the support–vector machine classifier cross–validation scheme. The obtained results show that the developed relevance analysis was capable of accurately differentiating normal, ictal and interictal activities.

Conclusions

The proposed approach provides the reliable identification of traces of interictal/ictal states of epilepsy. The introduced relevance rhythm diagrams of physiological rhythms provides effective means of monitoring epileptic seizures; additionally, these diagrams are easily implemented and provide simple clinical interpretation. The developed variability–based relevance analysis can be translated to other monitoring applications involving time–variant biomedical data.

Performance-power consumption tradeoff in wearable epilepsy monitoring systems

Automated seizure detection methods can be used to reduce time and costs associated with analyzing large volumes of ambulatory EEG recordings. These methods however have to rely on very complex, power hungry algorithms, implemented on the system backend, in order to achieve acceptable levels of accuracy. In size, and therefore power-constrained EEG systems, an alternative approach to the problem of data reduction is online data selection, in which simpler algorithms select potential epileptiform activity for discontinuous recording but accurate analysis is still left to a medical practitioner. Such a diagnostic decision support system would still provide doctors with information relevant for diagnosis while reducing the time taken to analyze the EEG. For wearable systems with limited power budgets, data selection algorithm must be of sufficiently low complexity in order to reduce the amount of data transmitted and the overall power consumption. In this paper, we present a low-power hardware implementation of an online epileptic seizure data selection algorithm with encryption and data transmission and demonstrate the tradeoffs between its accuracy and the overall system power consumption. We demonstrate that overall power savings by data selection can be achieved by transmitting less than 40% of the data. We also show a 29% power reduction when selecting and transmitting 94% of all seizure events and only 10% of background EEG.

Robust deep network with maximum correntropy criterion for seizure detection

Effective seizure detection from long-term EEG is highly important for seizure diagnosis. Existing methods usually design the feature and classifier individually, while little work has been done for the simultaneous optimization of the two parts. This work proposes a deep network to jointly learn a feature and a classifier so that they could help each other to make the whole system optimal. To deal with the challenge of the impulsive noises and outliers caused by EMG artifacts in EEG signals, we formulate a robust stacked autoencoder (R-SAE) as a part of the network to learn an effective feature. In R-SAE, the maximum correntropy criterion (MCC) is proposed to reduce the effect of noise/outliers. Unlike the mean square error (MSE), the output of the new kernel MCC increases more slowly than that of MSE when the input goes away from the center. Thus, the effect of those noises/outliers positioned far away from the center can be suppressed. The proposed method is evaluated on six patients of 33.6 hours of scalp EEG data. Our method achieves a sensitivity of 100% and a specificity of 99%, which is promising for clinical applications.

Feature extraction with stacked autoencoders for epileptic seizure detection

Scalp electroencephalogram (EEG), a recording of the brain's electrical activity, has been used to diagnose and detect epileptic seizures for a long time. However, most researchers have implemented seizure detectors by manually hand-engineering features from observed EEG data, and used them in seizure detection, which might not scale well to new patterns of seizures. In this paper, we investigate the possibility of utilising unsupervised feature learning, the recent development of deep learning, to automatically learn features from raw, unlabelled EEG data that are representative enough to be used in seizure detection. We develop patient-specific seizure detectors by using stacked autoencoders and logistic classifiers. A two-step training consisting of the greedy layer-wise and the global fine-tuning was used to train our detectors. The evaluation was performed by using labelled dataset from the CHB-MIT database, and the results showed that all of the test seizures were detected with a mean latency of 3.36 seconds, and a low false detection rate.

Automatic seizure detection based on the activity of a set of current dipoles: first steps

In this paper we show advantages of using an advanced montage scheme with respect to the performance of automatic seizure detection systems. The main goal is to find the best performing montage scheme for our automatic seizure detection system. The new virtual montage is a fix set of dipoles within the brain. The current density signals for these dipoles are derived from the scalp EEG signals based on a smart linear transformation. The reason for testing an alternative approach is that traditional montages (reference, bipolar) have some limitations, e.g. the detection performance depends on the choice of the reference electrode and an extraction of spatial information is often demanding. In this paper we explain the detailed setup of how to adapt a modern seizure detection system to use current density signals. Furthermore, we show results concerning the detection performance of different montage schemes and their combination.

Reliable epileptic seizure detection using an improved wavelet neural network

BACKGROUND

Electroencephalogram (EEG) signal analysis is indispensable in epilepsy diagnosis as it offers valuable insights for locating the abnormal distortions in the brain wave. However, visual interpretation of the massive amounts of EEG signals is time-consuming, and there is often inconsistent judgment between experts.

AIMS

This study proposes a novel and reliable seizure detection system, where the statistical features extracted from the discrete wavelet transform are used in conjunction with an improved wavelet neural network (WNN) to identify the occurrence of seizures.

METHOD

Experimental simulations were carried out on a well-known publicly available dataset, which was kindly provided by the Epilepsy Center, University of Bonn, Germany. The normal and epileptic EEG signals were first pre-processed using the discrete wavelet transform. Subsequently, a set of statistical features was extracted to train a WNNs-based classifier.

RESULTS

The study has two key findings. First, simulation results showed that the proposed improved WNNs-based classifier gave excellent predictive ability, where an overall classification accuracy of 98.87% was obtained. Second, by using the 10th and 90th percentiles of the absolute values of the wavelet coefficients, a better set of EEG features can be identified from the data, as the outliers are removed before any further downstream analysis.

CONCLUSION

The obtained high prediction accuracy demonstrated the feasibility of the proposed seizure detection scheme. It suggested the prospective implementation of the proposed method in developing a real time automated epileptic diagnostic system with fast and accurate response that could assist neurologists in the decision making process.

Automated quantification of spikes

Methods for rapid and objective quantification of interictal spikes in raw, unprocessed electroencephalogram (EEG) samples are scarce. We evaluated the accuracy of a tailored automated spike quantification algorithm. The automated quantification was compared with the quantification by two board-certified clinical neurophysiologists (gold-standard) in five steps: 1) accuracy in a single EEG channel (5 EEG samples), 2) accuracy in multiple EEG channels and across different stages of the sleep-wake cycles (75 EEG samples), 3) capacity to detect lateralization of spikes (6 EEG samples), 4) accuracy after application of a machine-learning mechanism (11 EEG samples), and 5) accuracy during wakefulness only (8 EEG samples). Our method was accurate during all stages of the sleep-wake cycle and improved after the application of the machine-learning mechanism. Spikes were correctly lateralized in all cases. Our automated method was accurate in quantifying and detecting the lateralization of interictal spikes in raw unprocessed EEG samples.

Compression-ratio-based seizure detection

For wireless seizure monitoring devices seizure detection and data compression are two critical tasks that need to be carefully designed against a very tight power budget to maximize the battery life. These two tasks are usually considered separately and algorithms for each are developed separately. In this paper, we consider having a single low-power algorithm for implementing both seizure detection and data compression. Towards that end, we investigated compression ratio (CR) as a seizure marker and show that the seizure detection can be achieved as a by-product of compression with no additional cost, and thus overall system power can be reduced. We show that the proposed method, the CR-based seizure detection has promising performance with 88% seizure detection accuracy, and 5.5 false positives per hour (FPh) without any computation overhead.

Electroencephalography in mesial temporal lobe epilepsy: a review

Electroencephalography (EEG) has an important role in the diagnosis and classification of epilepsy. It can provide information for predicting the response to antiseizure drugs and to identify the surgically remediable epilepsies. In temporal lobe epilepsy (TLE) seizures could originate in the medial or lateral neocortical temporal region, and many of these patients are refractory to medical treatment. However, majority of patients have had excellent results after surgery and this often relies on the EEG and magnetic resonance imaging (MRI) data in presurgical evaluation. If the scalp EEG data is insufficient or discordant, invasive EEG recording with placement of intracranial electrodes could identify the seizure focus prior to surgery. This paper highlights the general information regarding the use of EEG in epilepsy, EEG patterns resembling epileptiform discharges, and the interictal, ictal and postictal findings in mesial temporal lobe epilepsy using scalp and intracranial recordings prior to surgery. The utility of the automated seizure detection and computerized mathematical models for increasing yield of non-invasive localization is discussed. This paper also describes the sensitivity, specificity, and predictive value of EEG for seizure recurrence after withdrawal of medications following seizure freedom with medical and surgical therapy.

Optimal features for online seizure detection

This study identifies characteristic features in scalp EEG that simultaneously give the best discrimination between epileptic seizures and background EEG in minimally pre-processed scalp data; and have minimal computational complexity to be suitable for online, real-time analysis. The discriminative performance of 65 previously reported features has been evaluated in terms of sensitivity, specificity, area under the sensitivity-specificity curve (AUC), and relative computational complexity, on 47 seizures (split in 2,698 2 s sections) in over 172 h of scalp EEG from 24 adults. The best performing features are line length and relative power in the 12.5-25 Hz band. Relative power has a better seizure detection performance (AUC = 0.83; line length AUC = 0.77), but is calculated after the discrete wavelet transform and is thus more computationally complex. Hence, relative power achieves the best performance for offline detection, whilst line length would be preferable for online low complexity detection. These results, from the largest systematic study of seizure detection features, aid future researchers in selecting an optimal set of features when designing algorithms for both standard offline detection and new online low computational complexity detectors.

Model-based seizure detection for intracranial EEG recordings

This paper presents a novel model-based patient-specific method for automatic detection of seizures in the intracranial EEG recordings. The proposed method overcomes the complexities in the practical implementation of the patient-specific approach of seizure detection. The method builds a seizure model (set of basis functions) for a priori known seizure (the template seizure pattern), and uses the statistically optimal null filters as a building block for the detection of similar seizures. The process of modeling the template seizure is fully automatic. Overall, the detection method involves the segmentation of the template seizure pattern, rejection of the redundant and noisy segments, extraction of features from the segments to generate a set of models, selection of the best seizure model, and training of the classifier. The trained classifier is used to detect similar seizures in the remaining data. The resulting seizure detection method was evaluated on a total of 304 h of single-channel depth EEG recordings from 14 patients. The system performance is further compared to the Qu-Gotman patient-specific system using the same data. A significant improvement in the proposed system, in terms of specificity, is observed over the compared method.