Approximate Entropy-Based Epileptic EEG Detection Using Artificial Neural Networks

Multi-feature characterization of epileptic activity for construction of an automated internet-based annotated classification

Continuous monitoring of EEG is essential for the neurologist to detect the epileptic seizures that occur at various intervals. Since large volume of data need to be analyzed, visual analysis has been proven to be time consuming and subsequently automated detection techniques have gained importance in the recent years. For the biomedical research community, the major challenge lies in providing a solution to neurologists in terms of diagnosis and EEG database management. This paper discusses the automated detection of epileptic seizure using frequency domain and entropy parameters which helps in the construction of epileptic database for handling EEG data. Experimental study indicates that the suggested mode of operation can be used for internet based framework which contains pure epileptic patterns in the server. This can be retrieved and analyzed for detection and annotation of epileptic spikes in extensive EEG recordings.

Cross-conditional entropy and coherence analysis of pharmaco-EEG changes induced by alprazolam

RATIONALE

Quantitative analysis of electroencephalographic signals (EEG) and their interpretation constitute a helpful tool in the assessment of the bioavailability of psychoactive drugs in the brain. Furthermore, psychotropic drug groups have typical signatures which relate biochemical mechanisms with specific EEG changes.

OBJECTIVES

To analyze the pharmacological effect of a dose of alprazolam on the connectivity of the brain during wakefulness by means of linear and nonlinear approaches.

METHODS

EEG signals were recorded after alprazolam administration in a placebo-controlled crossover clinical trial. Nonlinear couplings assessed by means of corrected cross-conditional entropy were compared to linear couplings measured with the classical magnitude squared coherence.

RESULTS

Linear variables evidenced a statistically significant drug-induced decrease, whereas nonlinear variables showed significant increases. All changes were highly correlated to drug plasma concentrations. The spatial distribution of the observed connectivity changes clearly differed from a previous study: changes before and after the maximum drug effect were mainly observed over the anterior half of the scalp. Additionally, a new variable with very low computational cost was defined to evaluate nonlinear coupling. This is particularly interesting when all pairs of EEG channels are assessed as in this study.

CONCLUSIONS

Results showed that alprazolam induced changes in terms of uncoupling between regions of the scalp, with opposite trends depending on the variables: decrease in linear ones and increase in nonlinear features. Maps provided consistent information about the way brain changed in terms of connectivity being definitely necessary to evaluate separately linear and nonlinear interactions.

MULTISCALE ENTROPY ANALYSIS OF EEG RECORDINGS IN EPILEPTIC RATS

In this study, we investigate multiscale entropy (MSE) as a tool to evaluate the dynamic characteristics of electroencephalogram (EEG) during seizure-free, pre-seizure and seizure state, respectively, in epileptic rats. The results show that MSE method is able to reveal that EEG signals are more complex in seizure-free state than in seizure state, and can successfully distinguish among different seizure states. The classification ability of the MSE measures is tested using the linear discriminant analysis (LDA). Test results confirm that the classification accuracy of MSE method is superior to traditional single-scale entropy method. MSE method has potential in classifying the epileptic EEG signals.

A fuzzy logic system for seizure onset detection in intracranial EEG

We present a multistage fuzzy rule-based algorithm for epileptic seizure onset detection. Amplitude, frequency, and entropy-based features were extracted from intracranial electroencephalogram (iEEG) recordings and considered as the inputs for a fuzzy system. These features extracted from multichannel iEEG signals were combined using fuzzy algorithms both in feature domain and in spatial domain. Fuzzy rules were derived based on experts' knowledge and reasoning. An adaptive fuzzy subsystem was used for combining characteristics features extracted from iEEG. For the spatial combination, three channels from epileptogenic zone and one from remote zone were considered into another fuzzy subsystem. Finally, a threshold procedure was applied to the fuzzy output derived from the final fuzzy subsystem. The method was evaluated on iEEG datasets selected from Freiburg Seizure Prediction EEG (FSPEEG) database. A total of 112.45 hours of intracranial EEG recordings was selected from 20 patients having 56 seizures was used for the system performance evaluation. The overall sensitivity of 95.8% with false detection rate of 0.26 per hour and average detection latency of 15.8 seconds was achieved.

Automatic detection of seizure termination during electroconvulsive therapy using sample entropy of the electroencephalogram

Determining the exact duration of seizure activity is an important factor for predicting the efficacy of electroconvulsive therapy (ECT). In most cases, seizure duration is estimated manually by observing the electroencephalogram (EEG) waveform. In this article, we propose a method based on sample entropy (SampEn) that automatically detects the termination time of an ECT-induced seizure. SampEn decreases during seizure activity and has its smallest value at the boundary of seizure termination. SampEn reflects not only different states of regularity and complexity in the EEG but also changes in EEG amplitude before and after seizure activity. Using SampEn, we can more precisely determine seizure termination time and total seizure duration.

Classification of seizure and non-seizure EEG signals using empirical mode decomposition

In this paper, we present a new method for classification of electroencephalogram (EEG) signals using empirical mode decomposition (EMD) method. The intrinsic mode functions (IMFs) generated by EMD method can be considered as a set of amplitude and frequency modulated (AM-FM) signals. The Hilbert transformation of IMFs provides an analytic signal representation of the IMFs. The two bandwidths, namely amplitude modulation bandwidth (B(AM)) and frequency modulation bandwidth (B(FM)), computed from the analytic IMFs, have been used as an input to least squares support vector machine (LS-SVM) for classifying seizure and non-seizure EEG signals. The proposed method for classification of EEG signals based on the bandwidth features (B(A M) and B (FM)) and the LS-SVM has provided better classification accuracy than the method of Liang et. al [20]. The experimental results with the recorded EEG signals from a published dataset are included to show the effectiveness of the proposed method for EEG signal classification.

Quickest detection of drug-resistant seizures: an optimal control approach

Epilepsy affects 50 million people worldwide, and seizures in 30% of the cases remain drug resistant. This has increased interest in responsive neurostimulation, which is most effective when administered during seizure onset. We propose a novel framework for seizure onset detection that involves (i) constructing statistics from multichannel intracranial EEG (iEEG) to distinguish nonictal versus ictal states; (ii) modeling the dynamics of these statistics in each state and the state transitions; you can remove this word if there is no room. (iii) developing an optimal control-based "quickest detection" (QD) strategy to estimate the transition times from nonictal to ictal states from sequential iEEG measurements. The QD strategy minimizes a cost function of detection delay and false positive probability. The solution is a threshold that non-monotonically decreases over time and avoids responding to rare events that normally trigger false positives. We applied QD to four drug resistant epileptic patients (168 hour continuous recordings, 26-44 electrodes, 33 seizures) and achieved 100% sensitivity with low false positive rates (0.16 false positive/hour). This article is part of a Supplemental Special Issue entitled The Future of Automated Seizure Detection and Prediction.

A NEURAL NETWORK-BASED CLASSIFICATION MODEL FOR PARTIAL EPILEPSY BY EEG SIGNALS

Epilepsy is a disorder of cortical excitability and still an important medical problem. The correct diagnosis of a patient's epilepsy syndrome clarifies the choice of drug treatment and also allows an accurate assessment of prognosis in many cases. The aim of this study is to evaluate epileptic patients and classify subgroups of partial epilepsy by Multilayer Perceptron Neural Networks (MLPNNs). This is the first study to classify the partial epilepsy groups using the neural network according to EEG signals. 418 patients with epilepsy diagnoses according to International League against Epilepsy (ILAE, 1981) were included in this study. The epilepsy outpatients at the Neurology Department Clinic of Cukurova University Medical School between the years of 2002–2005 were examined and included in the study. The MLPNNs were trained by the parameters obtained from the EEG signals and clinical findings of the patients. Test results show that the MLPNN model is able to classify partial epilepsy with an accuracy of 91.5%. Moreover, new MLPNNs were constructed for determining significant variables on classification. The loss of consciousness in the course of seizure time variable caused the largest decrease in the classification accuracy when it was left out. In conclusion, we think that the classification performance of MLPNN model for partial epilepsy is satisfactory and this model may be used in clinical studies as a decision support tool to determine the partial epilepsy classification of the patients.

AUTOMATIC IDENTIFICATION OF EPILEPTIC EEG SIGNALS USING NONLINEAR PARAMETERS

Epilepsy is a brain disorder causing people to have recurring seizures. Electroencephalogram (EEG) is the electrical activity of the brain signals that can be used to diagnose the epilepsy. The EEG signal is highly nonlinear and nonstationary in nature and may contain indicators of current disease, or warnings about impending diseases. The chaotic measures like correlation dimension (CD), Hurst exponent (H), and approximate entropy (ApEn) can be used to characterize the signal. These features extracted can be used for automatic diagnosis of seizure onsets which would help the patients to take appropriate precautions. These nonlinear features have been reported to be a promising approach to differentiate among normal, pre-ictal (background), and epileptic EEG signals. In this work, these features were used to train both Gaussian mixture model (GMM) and support vector machine (SVM) classifiers. The performance of the two classifiers were evaluated using the receiver operating characteristics (ROC) curves. Our results show that the GMM classifier performed better with average classification efficiency of 95%, sensitivity and specificity of 92.22% and 100%, respectively.

Characterization of early partial seizure onset: frequency, complexity and entropy

OBJECTIVE

A clear classification of partial seizures onset features is not yet established. Complexity and entropy have been very widely used to describe dynamical systems, but a systematic evaluation of these measures to characterize partial seizures has never been performed.

METHODS

Eighteen different measures including power in frequency bands up to 300 Hz, Gabor atom density (GAD), Higuchi fractal dimension (HFD), Lempel-Ziv complexity, Shannon entropy, sample entropy, and permutation entropy, were selected to test sensitivity to partial seizure onset. Intracranial recordings from 45 patients with mesial temporal, neocortical temporal and neocortical extratemporal seizure foci were included (331 partial seizures).

RESULTS

GAD, Lempel-Ziv complexity, HFD, high frequency activity, and sample entropy were the most reliable measures to assess early seizure onset.

CONCLUSIONS

Increases in complexity and occurrence of high-frequency components appear to be commonly associated with early stages of partial seizure evolution from all regions. The type of measure (frequency-based, complexity or entropy) does not predict the efficiency of the method to detect seizure onset.

SIGNIFICANCE

Differences between measures such as GAD and HFD highlight the multimodal nature of partial seizure onsets. Improved methods for early seizure detection may be achieved from a better understanding of these underlying dynamics.

Evoked potentials in large-scale cortical networks elicited by TMS of the visual cortex

Single pulses of transcranial magnetic stimulation (TMS) result in distal and long-lasting oscillations, a finding directly challenging the virtual lesion hypothesis. Previous research supporting this finding has primarily come from stimulation of the motor cortex. We have used single-pulse TMS with simultaneous EEG to target seven brain regions, six of which belong to the visual system [left and right primary visual area V1, motion-sensitive human middle temporal cortex, and a ventral temporal region], as determined with functional MRI-guided neuronavigation, and a vertex "control" site to measure the network effects of the TMS pulse. We found the TMS-evoked potential (TMS-EP) over visual cortex consists mostly of site-dependent theta- and alphaband oscillations. These site-dependent oscillations extended beyond the stimulation site to functionally connected cortical regions and correspond to time windows where the EEG responses maximally diverge (40, 200, and 385 ms). Correlations revealed two site-independent oscillations ∼350 ms after the TMS pulse: a theta-band oscillation carried by the frontal cortex, and an alpha-band oscillation over parietal and frontal cortical regions. A manipulation of stimulation intensity at one stimulation site (right hemisphere V1-V3) revealed sensitivity to the stimulation intensity at different regions of cortex, evidence of intensity tuning in regions distal to the site of stimulation. Together these results suggest that a TMS pulse applied to the visual cortex has a complex effect on brain function, engaging multiple brain networks functionally connected to the visual system with both invariant and site-specific spatiotemporal dynamics. With this characterization of TMS, we propose an alternative to the virtual lesion hypothesis. Rather than a technique that simulates lesions, we propose TMS generates natural brain signals and engages functional networks.

PyEEG: an open source Python module for EEG/MEG feature extraction

Computer-aided diagnosis of neural diseases from EEG signals (or other physiological signals that can be treated as time series, e.g., MEG) is an emerging field that has gained much attention in past years. Extracting features is a key component in the analysis of EEG signals. In our previous works, we have implemented many EEG feature extraction functions in the Python programming language. As Python is gaining more ground in scientific computing, an open source Python module for extracting EEG features has the potential to save much time for computational neuroscientists. In this paper, we introduce PyEEG, an open source Python module for EEG feature extraction.

Performance of dynamic features in classifying scalp epileptic interictal and normal EEG

Over 50 million people worldwide suffer from epilepsy. Recently, researchers have proposed computer-aided epilepsy diagnostic systems based on classifying scalp epileptic interictal and normal EEG. Features used in the classification can be divided into two groups: classical spectral features and dynamic features. Classical spectral features are similar to major frequency component identification that physicians use in conventional EEG reading. Because dynamic features are new compared to classical spectral features, we are interested in knowing whether they are suitable for this classification problem. To study this, we build such a system and compare the results between using classical spectral features and dynamic features. Furthermore, we study which dynamic features are more suitable, i.e., more discriminative, by ranking them using F-score. According to the result, we discuss redesigning certain dynamic features for better classification. This research is a preliminary study of using dynamic features of scalp interictal EEG for epilepsy diagnosis.

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

A novel wavelet-based algorithm for real-time detection of epileptic seizures using scalp EEG is proposed. In a moving-window analysis, the EEG from each channel is decomposed by wavelet packet transform. Using wavelet coefficients from seizure and nonseizure references, a patient-specific measure is developed to quantify the separation between seizure and nonseizure states for the frequency range of 1-30 Hz. Utilizing this measure, a frequency band representing the maximum separation between the two states is determined and employed to develop a normalized index, called combined seizure index (CSI). CSI is derived for each epoch of every EEG channel based on both rhythmicity and relative energy of that epoch as well as consistency among different channels. Increasing significantly during ictal states, CSI is inspected using one-sided cumulative sum test to generate proper channel alarms. Analyzing alarms from all channels, a seizure alarm is finally generated. The algorithm was tested on scalp EEG recordings from 14 patients, totaling approximately 75.8 h with 63 seizures. Results revealed a high sensitivity of 90.5%, a false detection rate of 0.51 h(-1) and a median detection delay of 7 s. The algorithm could also lateralize the focus side for patients with temporal lobe epilepsy.

Analysis of amplitude-integrated EEG in the newborn based on approximate entropy

Amplitude-integrated electroencephalographic (aEEG), a cerebral-function-monitoring method, is widely used in response to the clinical needs for continuous EEG monitoring. In this paper, we present an approach to analyze aEEG in newborns based on approximate entropy (ApEn). Unlike the traditional aEEG signal processing and diagnosing methods, the Box-Cox transformation is substituted for semilogarithmic amplitude compression to keep the continuity of the signal, reduce the excessive compression of chaotic information in high amplitudes, and use ApEn, rather than the amplitudes of the borders, to estimate the degree of chaos in the signal. Experiments with aEEGs of 120 cases (32 normal and 88 abnormal of full-term infants, and 57 cases of preterm infants) were conducted to validate the effectiveness of the proposed method. The results show an aEEG signal analyzed based on the proposed algorithm always belongs to an abnormal case and needs to be examined by physicians if the corresponding indicator is considered abnormal. The novel description of aEEG could be helpful in detecting brain disorders in the newborn as a new clinical target.

Detecting complexity abnormalities in dyslexia measuring Approximate Entropy of electroencephalographic signals

Dyslexia constitutes a specific reading disability, a condition characterized by severe difficulty in the mastery of reading despite normal intelligence or adequate education. Electroencephalogram (EEG) signal may be able to play an important role in the diagnosis of dyslexia. The Approximate Entropy (ApEn) is a recently formulated statistical parameter used to quantify the regularity of a time series data of physiological signals. In this paper, we initially estimated the ApEn values in signals recorded from controls subjects and dyslectic children. These values were firstly used for the statistical analysis of the two groups and secondly as feature input in a classification scheme. We also used the cross-ApEn methodology to get a measure of the asynchrony of the signals recorded from different electrodes. This preliminary study provides promising results towards correct identification of dyslexic cases, analyzing the corresponding EEG signals.

Seizure detection in intracranial EEG using a fuzzy inference system

In this paper, we present a fuzzy rule-based system for the automatic detection of seizures in the intracranial EEG (IEEG) recordings. A total of 302.7 hours of the IEEG with 78 seizures, recorded from 21 patients aged between 10 and 47 years were used for the evaluation of the system. After preprocessing, temporal, spectral, and complexity features were extracted from the segmented IEEGs. The results were thresholded using the statistics of a reference window and integrated spatio-temporally using a fuzzy rule-based decision making system. The system yielded a sensitivity of 98.7%, a false detection rate of 0.27/h, and an average detection latency of 11 s. The results from the automatic system correlate well with the visual analysis of the seizures by the expert. This system may serve as a good seizure detection tool for monitoring long-term IEEG with relatively high sensitivity and low false detection rate.

Automated epilepsy diagnosis using interictal scalp EEG

Over 50 million people worldwide suffer from epilepsy. Traditional diagnosis of epilepsy relies on tedious visual screening by highly trained clinicians from lengthy EEG recording that contains the presence of seizure (ictal) activities. Nowadays, there are many automatic systems that can recognize seizure-related EEG signals to help the diagnosis. However, it is very costly and inconvenient to obtain long-term EEG data with seizure activities, especially in areas short of medical resources. We demonstrate in this paper that we can use the interictal scalp EEG data, which is much easier to collect than the ictal data, to automatically diagnose whether a person is epileptic. In our automated EEG recognition system, we extract three classes of features from the EEG data and build Probabilistic Neural Networks (PNNs) fed with these features. We optimize the feature extraction parameters and combine these PNNs through a voting mechanism. As a result, our system achieves an impressive 94.07% accuracy.

Drug effect on EEG connectivity assessed by linear and nonlinear couplings

Quantitative analysis of human electroencephalogram (EEG) is a valuable method for evaluating psychopharmacological agents. Although the effects of different drug classes on EEG spectra are already known, interactions between brain locations remain unclear. In this work, cross mutual information function and appropriate surrogate data were applied to assess linear and nonlinear couplings between EEG signals. The main goal was to evaluate the pharmacological effects of alprazolam on brain connectivity during wakefulness in healthy volunteers using a cross-over, placebo-controlled design. Eighty-five pairs of EEG leads were selected for the analysis, and connectivity was evaluated inside anterior, central, and posterior zones of the scalp. Connectivity between these zones and interhemispheric connectivity were also measured. Results showed that alprazolam induced significant changes in EEG connectivity in terms of information transfer in comparison with placebo. Trends were opposite depending on the statistical characteristics: decreases in linear connectivity and increases in nonlinear couplings. These effects were generally spread over the entire scalp. Linear changes were negatively correlated, and nonlinear changes were positively correlated with drug plasma concentrations; the latter showed higher correlation coefficients. The use of both linear and nonlinear approaches revealed the importance of assessing changes in EEG connectivity as this can provide interesting information about psychopharmacological effects.

A closed-loop brain computer interface for real-time seizure detection and control

The worldwide prevalence of epilepsy is approximately 1%, and 25% of epilepsy patients cannot be treated sufficiently by available therapies. Brain stimulation with closed-loop seizure control has recently been proposed as an innovative and effective alternative. In this paper, a portable closed-loop brain computer interface for seizure control was developed and shown with several aspects of advantages, including high seizure detection rate (92-99% during wake-sleep states), low false detection rate (1.2-2.5%), and small size. The seizure detection and electrical stimulation latency was not greater than 0.6 s after seizure onset. A wireless communication feature also provided flexibility for subjects freeing from the hassle of wires. Experimental data from freely moving rats supported the functional possibility of a real-time closed-loop seizure controller.

Real-time differentiation of nonconvulsive status epilepticus from other encephalopathies using quantitative EEG analysis: a pilot study

PURPOSE

Distinguishing nonconvulsive status epilepticus (NCSE) from some nonepileptic encephalopathies is a challenging problem. In many situations, NCSE and nonepileptic encephalopathies are indistinguishable by clinical symptoms and can produce very similar electroencephalography (EEG) patterns. Misdiagnosis or delay to diagnosis of NCSE may increase the rate of morbidity and mortality.

METHODS

We developed a fast-differentiating algorithm using quantitative EEG analysis to distinguish NCSE patients from patients with toxic/metabolic encephalopathy (TME). EEG recordings were collected from 11 patients, including 6 with NCSE and 5 with TME. Three nonlinear dynamic measures were used in the proposed algorithm: the maximum short-term Lyapunov exponent (STLmax), phase of attractor (phase/angular frequency), and approximate entropy (ApEn). A further refined metric derived from STLmax and phase of attractor (the mean distance to EEG epoch samples from their centroid in the feature space) was also utilized as a criterion. Paired t tests were carried out to further clarify the separation between the EEG patterns of NCSE and TME.

RESULTS

Computational results showed that the performance of the proposed algorithm was sufficient to distinguish NCSE from TME. The results were consistent in all subjects in our study.

CONCLUSIONS

The study presents evidence that the maximum short-term Lyapunov exponents (STLmax) and phase of attractors (phase/angular frequency) can be useful in assisting clinical diagnosis of NCSE. Findings presented in this article provide a promising indication that the proposed algorithm may correctly distinguish NCSE from TME. Although the exact mechanism of this association remains unknown, the authors suggest that epileptic activity is highly associated with and can be modeled by dynamic systems.

A fuzzy rule-based system for epileptic seizure detection in intracranial EEG

OBJECTIVE

We present a method for automatic detection of seizures in intracranial EEG recordings from patients suffering from medically intractable focal epilepsy.

METHODS

We designed a fuzzy rule-based seizure detection system based on knowledge obtained from experts' reasoning. Temporal, spectral, and complexity features were extracted from IEEG segments, and spatio-temporally integrated using the fuzzy rule-based system for seizure detection. A total of 302.7h of intracranial EEG recordings from 21 patients having 78 seizures was used for evaluation of the system.

RESULTS

The system yielded a sensitivity of 98.7%, a false detection rate of 0.27/h, and an average detection latency of 11s. There was only one missed seizure. Most of false detections were caused by high-amplitude rhythmic activities. The results from the system correlate well with those from expert visual analysis.

CONCLUSION

The fuzzy rule-based seizure detection system enabled us to deal with imprecise boundaries between interictal and ictal IEEG patterns.

SIGNIFICANCE

This system may serve as a good seizure detection tool with high sensitivity and low false detection rate for monitoring long-term IEEG.

Seizure detection in temporal lobe epileptic EEGs using the best basis wavelet functions

In this paper, we propose a novel method using best basis wavelet functions and double thresholding that are well suited for detecting and localization of important epileptic events from noisy recorded seizure EEG signals. Our technique is based on dyadic wavelet decomposition and is mainly concerned detection of single epileptic transients within the observation sequence, such as ictal and interictal epochs of EEG. In our experiment we use temporal lobe epileptic data recorded during 84 h from four patients diagnosed with epilepsy. We have achieved promising results that demonstrate efficiency and simplicity that can be used in clinical studies as an automatic decision support tool. Thus reduce the physician's workload and provide accurate diagnosis of epileptic seizures.

Entropy analysis on EEG signal in a case study of focal myoclonus

Electrophysiological studies provide useful information for diagnosis and classification of myoclonus, and for the investigation of its generative mechanisms, due to association of myoclonus with abnormally increased excitability of cortical structures. In this work we analyzed the polygraphic data of a 7-year old girl affected by continuous partial epilepsy with focal myoclonus both related and not related with epileptiform discharges on EEG. We applied Sample Entropy (SampEn) and Lempel-Ziv complexity (LZ) methods to investigate the regularity and complexity content of EEG recordings and to find possible analogies in the behaviour of non-parametric complexity measures in epilepsy and in myoclonus. Our results show that these algorithms succeeded in finding a significant difference between the hypothesized focus on C3 electrode and the contralateral electrode C4, for EEG correlated myoclonus. A significant difference between the two contralateral electrodes (C3-C4) was also found for non EEG correlated myoclonus, but only by means of SampEn. This preliminary study confirmed the ability of entropic methods in discriminating myoclonic events. Indeed, near the myoclonic focus location both SampEn and LZ methods showed below average values.