Search

Objective : This study demonstrates how functional connectivity (FC) patterns are affected in direct relation to the lobe that is mostly affected by seizures. Methods : The novel idea of penalized FC (pFC) maps is compared against standard FC maps in the four fundamental EEG frequency sub-bands. The FC measure between any two specific electrodes is scaled depending on the probability of true FC between them, and their power content with respect to the two electrodes of maximum power within each frequency sub-band. The algorithm is automated and introduces adaptive power penalization based on the power distribution of the different sub-bands. Results : The pFC maps were found to be more effective at suppressing the local connectivity in the lobes that are less affected by the interictal epileptiform discharges (IEDs). More precisely, given the least amount of power penalization, pFC maps of the theta sub-band reveal statistical significance in terms of increased local connectivity margin of the affected region as compared to the standard FC maps. However, they cannot be solely relied upon as other sub-bands could alternatively show high local connectivity across different patients within the region of interest. Conclusion : penalized functional connectivity maps intrinsically provide more information regarding the whole brain network in context to regions of interest where the active lobe is determined by the neurologists to contain the focal source. Significance : Findings suggest that (1) the significant sub-band varies from patient to patient while remaining relatively consistent within the IED segments of a same patient, and (2) the pFC maps have an advanced capability in terms of pinpointing to a region of interest of the active lobe, and as such can play a critical role in providing insight as to a region of interest where the 3D source might be located when solving the ill-posed inverse problem.

Objective: Stereoelectroencephalogram (SEEG) has been widely adapted to detect the electrical activity of patients with epilepsy. Due to the low-quality, large-amount, high-dimensionality characteristics of SEEG data, it is still challenging to comprehensively employ the SEEG signals to automatically and precisely determine the seizure onset zone (SOZ). This is because there is lack of an effective criterion for clinicians to select the target electrodes, which is of great importance for SOZ localization. Methods: We propose a SOZ localization method via analyzing the long-term SEEG monitoring for preoperative planning of epilepsy surgery. Considering that high frequency oscillations can reflect physiological brain activity of epileptic patients, we first extract the high-frequency features of the SEEG signals and utilize the convolutional neural network (CNN) to detect the interictal and seizure segments. Then we propose a novel criterion, namely adaptive high frequency epileptogenicity index (AHFEI), to determine the target electrodes. Results: We compare our SEEG-determined target electrodes with three preoperative planning of successful focal epilepsy resective surgery cases, finding that most localization results of our method are in consistent with clinical successful decision making, while the performance of our method outperforms than the state-of-the-art method for SOZ localization. Conclusion: Our SEEG-determined SOZ localization method can assist clinicians in preliminarily selecting the potential target electrodes according to long-term SEEG data automatically and effectively. Clinical Impact: The proposed automatic SOZ localization method has achieved satisfactory performance in the preliminary study, which has the great potential to be integrated into function-structure fused clinical decision-making system.

People with epilepsy will benefit from the adoption of the Brain Computer Interface (BCI) for seizure prediction. Epilepsy is a frequent neurological condition of the brain's Central Nervous System that has negative consequences for patients. The human brain has millions of neurons, each of which is responsible for a distinct motor action in the human body. The neurons send out aberrant electrical signals that go throughout the body, causing involuntary movements and seizures. During the seizure, the patient collapses and experiences jerky movements, tongue biting, nausea, and other symptoms. Recurrent seizures are a hallmark of epilepsy patients. EEG (Electroencephalogram) is one of the common procedures to observe the signals produced by brain and helps to examine seizure occurrence in the patients. Seizure has four different stages ictal (during seizure), pre-ictal (before seizure), inter ictal (time interval between two successive seizures) and post ictal (after seizure) (after seizure). The patient will be kept in a safer state if the pre-ictal stage of the patient can be predicted. With the introduction of cutting-edge Neural Network techniques, their various activation functions, and Gradient Boosting, the patient's pre-ictal state can be predicted and diagnosed, allowing proper seizure medication to be administered. The suggested model uses the Neural Network Based Ramp Raise Optimized Algorithm (NNBRROA) with hyper parameter tuning to predict pre-ictal patients with 98 percent accuracy.

We present a framework that allows to investigate the fine-grained patterns of interictal spikes present in magne-toencephalographic (MEG) data of epileptic patients. This analysis could provide a valuable information regarding epileptogenic zone (EZ) location. We use a traveling wave model to regularize the MEG inverse problem. Our algorithm represents spike propagation patterns as a superposition of local waves traveling along radial paths stemming from a single origin. Using the positively constrained LASSO technique we scan over the wave onset moment and propagation velocity parameters to determine their combination that yields the best fit to the MEG sensor data of each spike. We use realistically simulated MEG data to validate the algorithm ability to successfully track interictal activity on a millimeter-millisecond scale. Also we examine MEG data from three patients with drug-resistant epilepsy. Wave-like spike patterns with clear propagation dynamics are found in a fraction of spikes, whereas the other fraction can not be explained by the wave propagation model with a small number of propagation directions. Moreover, the spike waves with clear propagation dynamics exhibit spatial segregation and match the clinical records on EZ availabel for two patients out of three.

Epilepsy is a common neurological disorder characterized by recurrent epileptic seizures. These seizures have different intensities and might lead to accidents or, in the worst case, to sudden death. Therefore, being able to predict epileptic seizures would allow patients to be prepared, reducing the risk of injury. This paper focuses on epileptic seizure prediction using EEG (Electroencephalogram) signals. In contrast to the standard approach where the preictal state is assumed to have a constant duration in all the seizures of a patient, we propose a new method that labels each seizure individually exploiting clustering. Our labeling approach, which was applicable for 38% of the selected seizures, results in substantial improvements compared to the standard one. In fact, it reduces noise in the labels and improves the performance of the binary classifier used to distinguish the interictal and preictal states. Hence, our results suggest that the preictal duration is seizure-specific, not only patient-specific. Finally, we show that our method is able to predict 17 out of 18 (94%) seizures between 15 and 85 minutes, before seizure onset.

Epilepsy is one of the most common chronic neurological disorders, affecting approximately 50 million people worldwide according to the World Health Organization. This disorder mainly presents four kinds of events: pre-ictal, ictal, post-ictal, and inter-ictal. Epilepsy can be diagnosed through an electroencephalogram (EEG). Inter-ictal activity is one of the widely accepted epilepsy symptoms on an EEG. However, the differentiation between normal and inter-ictal EEG segments is difficult because they can have similar patterns. Also, EEG from patients with epilepsy can contain normal events. In this work, we built classifiers to differentiate between normal, ictal, and inter-ictal EEG. Using Discrete Wavelet Transform multilevel decomposition of the signal is done. At each stage, the vital features are collected from the approximation and detailed coefficients belonging to a certain frequency range where the epilepsy is identifiable. Some of the features are directly extracted from the EEG signal. The machine learning algorithm is used to train and test the wide range of classifiers that suits the signal. This proposed method is implemented and tested with 98 percent accuracy. Here an emergency mail is sent to the doctor if any abnormality is found in the EEG signal using the Internet of Things (IoT).

This work proposes an automated epileptic seizure detection pipeline using generated rhythmicity spectrograms and Generic Convolutional Neural Networks (CNN). 1D multi-channel, scalp-EEG signals taken from a publically available EEG database (CHB-MIT Scalp EEG database, version: 1.0.0) were converted to time-variable, non-overlapped and one-sided rhythmicity spectrograms (2D images) through Short-time Fourier transform (STFT) method for patient no. chb01, chb02, chb03 and chb05. A two class, supervised classification between seizure (ictal) and non-seizure images was performed. Thorough cross-patient test set analysis has been presented along with evaluation metrics such as precision, recall, F1-score, and loss and accuracy of the model on the test set. The generic model achieved an average training, validation and test set accuracy up-to 91.89, 88.17 and 61% respectively. An automated epileptic seizure detection system can escalate the process of diagnosis and early decision to surgery which may aid in quality of life (QOL) of diseased patients.

Kocaeli University;Kocaeli University Technopark 2021 International Conference on INnovations in Intelligent SysTems and Applications, INISTA 2021 -- 25 August 2021 through 27 August 2021 -- 172175 The development of systems that can predict epilepsy seizures in real time offers great hope for epilepsy patients. These systems aim to prevent accidents that patients may experience due to loss of consciousness during seizures. Therefore, systems that can predict epileptic seizures should both work in real time and be designed to maintain the daily activities of the patient. In this case, a system with as few electrodes as possible should be developed. In this study, it is aimed to choose the most appropriate electrode in predicting epileptic seizures. Channel selection is made according to two parameters and its effect on seizure prediction is examined. The first parameter is the difference in variance between preictal and interictal; The other parameter is the weighted average sensitivity (WAS). The Rusboosted Tree ensemble classification is used to calculate WAS. The prediction process is carried out with the method we proposed in the previous study. For performance evaluation, prediction accuracy, sensitivity (SEN) and false alarm rates per hour (FPR) are calculated. The prediction performance for the channel selected according to the variance difference results are 69%, 70.9% and 0.054 respectively and the for the channel selected according to WAS results are 69%, 71.8% and 0.031 respectively. © 2021 IEEE. Kocaeli Üniversitesi: :2018/063 ACKNOWLEDGMENT This work was supported by Kocaeli University, Scientific Research Projects Coordination Unit, under project number:2018/063.

Epilepsy is a severe neurological disease that can result in life threatening situations. Therefore, accurate, reliable, and robust epilepsy diagnostic systems is of high importance. The design of such systems involves feature extraction and machine learning (ML) models. Thus, we introduce a novel Self-Similarity Index (SSI) feature based on Hjorth complexity and Katz fractal dimension. SSI is explored individually and in combination with other features extracted from both raw EEG signal and Wavelet Package Decomposition (WPD) coefficients. Several ML models are tested on the Bonn dataset to identify the most suitable model for classifying data into 3 classes - healthy, ictal, and inter-ictal. We show that the proposed SSI is superior to other features when used in a single-feature Support Vector Machine (SVM) classifier and yields 81.80% accuracy. This result is further improved to 99.00% when SSI is combined with 24 other commonly extracted features. Thus, SSI is a promising feature that allows for high classification accuracy with limited data and number of channels.

Interictal epileptiform discharges (IEDs) can have various morphologies as well as spatial distributions and sometimes are associated with other brain activities, resulting in uncertainty in their labeling. Such an uncertainty corresponds to the probability of a waveform being an IED. Here, we incorporate this probability in an IED detection system which combines spatial component analysis (SCA) with the IED probabilities referred to as SCA-IEDP-based method. For comparison, we also propose SCA-based method in which the probability of being IED is ignored. The outcome shows that the SCA-IEDP outperforms SCA.

Epilepsy is one of the most common neurological disorders generally characterized by sudden and recurrent unprovoked seizures, which is commonly diagnosed through visual inspection of electroencephalography (EEG) signals. However, since EEG signals are highly complex, nonlinear and nonstationary in nature, both visual inspection and the existing computed aid detection approaches fail to capture the intrinsic dynamics of seizure events, leading to unsatisfactory detection performance. Therefore, an accurate computer-aided automatic epileptic seizure diagnosis system is valuable and urgently needed in clinical application. In this study, we propose an automatic classification method using nonlinear dynamics features and nonlinear classifiers for determining whether the EEG signals are in interictal or ictal activities, respectively. The main novelty of this paper is that investigating the performance of discernable entropy-based nonlinear dynamics features used for automatic diagnosis of an epileptic seizure. First, the discrete wavelet transform is applied to obtain de-noised and sub-band signals. Second, the nonlinear dynamics features are extracted based on information entropy theory. Subsequently, one-way analysis of variance and forward sequential feature selection technique are employed to identify the most explainable features for seizure detection. Then the selected features are fed to six different classifiers, namely, least square-support vector machine (LS-SVM), K-nearest neighbors, logistic regression, linear discriminant analysis, Naive Bayes classifier, and random forest, for automatic diagnosis of the epileptic seizure. Finally, a public EEG dataset from Bonn University, Germany, is utilized to investigate the performance of the proposed method. The experimental results show that the LS-SVM yields the optimal performance with an accuracy of 99.50%, a sensitivity of 100.00%, and a specificity of 99.40%. Our proposed method shows great potential for real-time diagnosis of the epileptic seizures, which can be deployed in hospitals to assist clinicians in accurate diagnosis of seizures.

Epilepsy is a chronic disorder that causes unprovoked, recurrent sudden abnormal reactions of the brain. Characterizing electroencephalogram (EEG) signals of the patient is an effective way for the early prediction of epileptic seizures. In this paper, a new method called the entropy of visibility heights of hierarchical neighbors (EVHHN) is proposed to detect seizures from the EEG signals. First, the visibility relationships of three nearest neighbors are determined by a visibility criterion. Then, we compute the visibility heights of three nearest neighbors for each data point. Next, the four different kinds of entropy associated with neighbor visibility states are calculated to characterize the EEG signals and finally these features are validated by LS-SVM classifiers. In the experiment, the normal and ictal EEG signals are classified with the accuracy of 99.6%, meanwhile, the interictal and ictal EEG signals are distinguished with the accuracy of 98.35%, which proves the effectiveness of our proposed method. Notably, the computational time of extracting features for each set is 1.751 s, which is largely reduced compared with other weighted visibility graph-based methods. In conclusion, the EVHHN can potentially be an effective method for characterizing complicated EEG signals and real-time detection of epileptic seizures.

Epilepsy is the most unpredictable and recurrent disease among neurological diseases. Early detection of epileptic seizures can play a critical role in providing timely treatment to patients especially when a patient is in a remote area. This paper uses deep learning framework to detect epilepsy in the Electroencephalography (EEG) signal. The dataset used is publicly available and has a recording of three kinds of EEG signals: pre-ictal, inter-ictal (seizure-free epileptic) and ictal (epileptic with seizure). The proposed Long Short-Term Memory (LSTM) classifier classifies these three kinds of signals with up to 95% accuracy. For binary classification such as detection of inter-ictal or ictal only, its accuracy increases to 98%. The EEG signal is modelled as wide sense non-stationary random signal. Hurst Exponent and Auto-regressive Moving Average (ARMA) features are extracted from each signal. In this work, two different configurations of LSTM architecture: single-layered memory units and double-layered memory units are also modelled. After standardising the features, double-layered LSTM approach gives the highest accuracy in comparison to previously used Support Vector Machine (SVM) classifier and proved to be computationally efficient at Graphics Processing Unit (GPU).

EEG signal records the nerve activity in the brain, which is of great significance for the diagnosis and treatment of epilepsy. Effective automatic diagnosis method for epilepsy interictal period and ictal period can predict epilepsy and prevent the hurt to the body. In this paper, an automatic epilepsy detection method is proposed based on support vector machine classifier which use the sample entropy and standard deviation features selected by the reliefF algorithm from the components of EEG signals using empirical mode decomposition and variational mode decomposition. The epilepsy EEG database of Bonn University is used to evaluate the method. The experimental results show that proposed method can distinguish the epilepsy EEG signal between interictal period and ictal period in terms of sensitivity, specificity, precision, and accuracy. The best classification accuracy is up to 97.00% using support vector machine classifier with fine gaussian kernel function based on selected features.

Interictal 18F-fluorodeoxyglucose-PET (18F-FDG PET) is widely used in presurgical evaluation for epileptogenic zone (EZ) identification in patients with temporal lobe epilepsy. However, the area of PET hypometabolism is usually more extensive than the epileptogenic foci, providing approximate localization rather than precise. In this study, we evaluated the performance of 18F-FDG PET for classifying the EZs and the non-EZs using the radiomics features in addition to the hypometabolic features. The PET images of twelve patients with temporal lobe epilepsy were segmented into small cubes, which were labelled as EZs defined by the surgical resection areas or non-EZs. After extracting the radiomics features as well as the hypometabolic features of each cube, a support vector machine classifier was employed to classify each cube into EZ or non-EZ. The classifier performed well with the area under the curve (AUC) of $0.893 \pm 0.051$ , and using the combined features together got better performance than using the hypometabolic features only. The findings of this study merit further explorations to encourage the use of radiomics features in clinical decision for the EZ localization based on 18F-FDG PET.

Interictal epileptiform discharges (IEDs), or “spikes”, are indicative of seizures and are useful in the diagnosis of epilepsy. Automated algorithms like machine learning models have shown promise in detecting spikes without requiring human supervision. However, there is still a lack of comprehensive, high-quality, and accessible data that can be used to train these algorithms. The goal of this work is to assess whether Generative Adversarial Networks (GANs) can create synthetic spikes that are both realistic and that can improve the performance of machine learning classification. Here we developed an Auxiliary Classifier GAN (AC-GAN) to generate synthetic spikes and trained it on expert-annotated real spike events from intracranial EEG recordings. We found that the AC-GAN was capable of successfully generating realistic synthetic spikes, demonstrated by qualitative and quantitative metrics. Furthermore, the synthetic spikes improved classification accuracy when augmented to the training data. Our findings exemplify the utility of GANs in assisting with epilepsy diagnosis through synthesis of realistic neurophysiological signal waveforms and through data augmentation in classification.

Predicting epilepsy ahead of its occurrence has been an arduous job for scientists for a long time. Epileptic patients are still endeavoring to find a prosperous way to evade seizures to improve the quality of their lives. In this paper, we propose a novel deep learning system for epileptic seizure prediction using multi-channel electroencephalogram (EEG) recordings from the scalp of human brains. The proposed system is patient-specific and is predicated on the classification between the interictal and preictal brain states for the epileptic patient. The system uses a two-dimensional convolutional variational autoencoder and trains it once in a supervised way for automatic feature learning and classification. Within a prediction window of up to one hour, our proposed system achieved an average sensitivity of 94.45% and 0.06FP/h average false prediction rate which makes it one of the most efficient among state-of-the-art methods.

Epilepsy is categorized as a seizure disorder that causes unpredictable chronic seizures. Electroencephalogram (EEG) records electrical activities of brain that can be used to detect epileptic seizures. The EEG data of an epileptic patient includes a vast number of EEG details that may involve timeful manual interpretations. Automatic EEG signal interpretation with sophisticated signal processing techniques plays a major role in detecting epilepsy from EEG recordings. In this study, a new approach of feature extraction for forecasting onset of epileptic seizures is introduced. For analytical processing, the EEG signals are decomposed using Tunable Q Factor Wavelet Transformation. The Statistical Features (SF), Temporal Features (TF) and Global Features (GF) for the sub-bands are extracted using Auto Encoders, Continuous Bag of Words (CBoW) for Signals and Principal Component Analysis (PCA). The EEG signals classification is performed using ANN classifier to distinguish between the normal, inter-ictal and ictal classes.

Electroencephalograph (EEG) is a record of electrical activity of human brain signal to represent the movement, various analysts are functioning on brain as they are interested with the possibility of mystery, consideration and sense from the outside and inward upgrades. Feature extraction, examination, and grouping of electroencephalogram signals areas yet testing issue for scientists because of the varieties of the EEG signals. Various elements are utilized to recognize seizures, unconsciousness, and encephalopathy. And mind passing, and so forth. In any case, we include seen so as to extricated elements from equal sorts of signal changes have not compelling to separate the seizures time frames as well as after a seizure and between seizures, of electrical of activity of the brain signal. Here implementing another methodology for feature extraction utilizing high recurrence parts from change. We additionally consolidate the new component with the transmission capacity include removed since the experimental method decomposition Empirical Mode Decomposition. These are highlighted characteristic have been utilized since a contribution to classification method to arrange Ictal and Interictal period of seizures of brain activity of the signals from various brain areas. Exploratory outcomes illustrate that the projected strategy beats the current condition of threat technique for improved categorization of during a seizure, between seizures time phase of brain disorder epilepsy used for target dataset.

To aid doctors with the process of scanning long-term Electroencephalogram (EEG) records for seizure detection, automatic seizure detectors (ASD) have been proposed. ASD have many clinical uses and have gained much recent interest in the past few decades. However, due to high computational loads of the ASD algorithms; their real-time application has been hindered. The aim of this work is to build a detector that has low complexity and can detect seizures with high sensitivity and minimum latency. To this end, this paper proposes a patient-specific seizure onset detector that uses a single feature based on neural synchrony as it has been shown that once the brain approaches ictal activity, neural activity becomes less chaotic and more synchronous. Leveraging on this phenomena, the condition number is utilized to measure the synchronization between EEG channels. The developed detector has three main stages: preprocessing, condition number calculation and classification. After filtering an input EEG record, the condition number of an EEG window is computed and fed into a classifier that determines whether the current window is normal or abnormal. Classification was done by two forms of a support vector machine (SVM) classifier; namely, binary SVM and a one-class SVM. The developed detector was evaluated using the CHB-MIT dataset and was evaluated using 10 patients from this dataset. The proposed detector achieved a sensitivity of 97% and a maximum false positive rate of five false alarms per hour. The performance of the detector is comparable to seizure detectors that use multiple features; but lightweight for use for real-time processing of EEG data.

Epilepsy is a neurological disorder that affects a wide population of the world. It is categorized by a number of repeated seizures. The electroencephalogram (EEG) has been a most powerful tool in the diagnosis of epilepsy. This work proposes a long short-term memory (LSTM) network with only one feature, characterizing cross-frequency correlation, for the classification of different EEG states, viz., ictal (during seizure), interictal (between seizures) and preictal (before seizure). The LSTM network has been assessed using open CHB-MIT scalp EEG database consisting of 983 hours’ long-term EEG recordings. The accuracy of 95.71% was achieved for preictal duration of 30 minutes (mins) and when EEG of 1 hour (hr) before and after the seizures has not been included in interictal state. The results of the evaluation suggest the use of proposed network with minimal set of features in seizure detection as well as seizure prediction.

Reliable epileptic seizure prediction is significantly important in improving the life of patients and enhancing the therapy effect. In this paper, a novel seizure prediction algorithm is proposed employing the tensor decomposition on long-term intracranial EEG recordings. The modified Stockwell transform (MST) is conducted on the segmented EEG signals to transform into two-dimensional instantaneous power spectra. Then, the third-order tensor representation of the multi-channel EEG signals are structured with the models of time, frequency and space. Tucker decomposition, one valid tensor decomposition method, is applied to obtain the principal components of the EEG tensors and the smaller core tensors after decomposition are extracted as features of interictal EEG and preictal EEG. After that, the classification of preictal and interictal data is achieved by feeding the features into Bayesian Linear Discriminant Analysis (BLDA) classifier. The evaluation of the proposed algorithm is carried out on the Freiburg EEG database and a sensitivity of 88.49% for the seizure occurrence period of 30 min, meanwhile, a sensitivity of 97.62% for the seizure occurrence period of 50 min are yielded with a false alarm rate of 0. 25/h. The results show that this algorithm based on tensor analysis has notable performance for seizure prediction.

Electroencephalography reflects the averaged electrical activity of neurons associated with physiological and neural disease processes, i.e., epilepsy. In this study, EEG signals recorded during a complex partial seizure are modeled using a coupled Duffing-van der Pol oscillator. The purpose of the paper is the statistical analysis of parameter variability during three phases of seizure: pre-, ictal, and post-ictal. In order to determine the impact of intra -variability on parameter values during seizure spread, in the study, signals from 10 patients are analyzed. Our data are reliable in terms of each patient's sex, age, and medical history. For the first time, the results of the optimal values model parameter are explained based on physiological phenomena. Non-linear models' parameters may be useful for pre-and post-ictal sequences differentiation.

Epilepsy is the second most popular neurological disorder affecting 65 million people around the world. Seizures are classified into two kinds; focal and generalized ictal activities, reflecting the spread of seizure activity on the brain. Focal seizures start and affect specific regions of the brain, whereas the generalized propagates throughout the brain. Current approaches to developing an automatic seizure detection algorithm do not consider the types of seizures. However, to detect the focal seizures, the locations of onset of seizure must be identified by an expert through inspection of the electroencephalogram (EEG), which is an expensive and time-consuming procedure. Moreover, most proposed methods are patient-specific and cannot be generalized on an unseen patient, limiting the clinical usage of previous studies. This work presents a generalizable seizure detection algorithm by considering different seizure types. After pre-processing data and rejecting artifacts, a deep neural network is used to extract robust representations across seizures and a population. The proposed method includes deep recurrent and convolutional neural networks to capture spatial and temporal information simultaneously. Experiments on the TUH EEG seizure dataset, which contains both generalized and focal seizures, show that the proposed method increases the accuracy over state-of-the-art from 80.72% to 82%, precision from 67.55% to 71.69%, and sensitivity from 80% to 85%.

Using electroencephalography (EEG) data from epileptic patients1, we investigated and compared functional connectivity networks of three various types of epileptiform discharges (ED; single, complex & repetitive spikes) in 4 regions of the brain. Our results showed different connectivity patterns among three ED types within-and between-brain regions. The one-way ANOVA test indicated significant differences between the mean of the average connectivity matrices (ACMs) of the single spike, which characterize focal epilepsy, and the other two ED types (complex & repetitive) which characterize generalized epilepsy. The interictal EEG segments, through the connectivity patterns they yield, could be considered as one of the key indicators for the diagnosis of focal or generalized epilepsy.

Epilepsy is a neural disease caused by brain signal abnormalities. There are three phases in epilepsy, the pre-ictal, ictal, and post-ictal phase. To distinguish those phases, usually EEG signal is used. However, there is a study mentioning the connection between epilepsy and heart signals, so there is a probability to distinguish those phases using ECG. This study is made for distinguish the three phases in epilepsy and the normal condition of epilepsy patient using K Nearest Neighbors (KNN) algorithm. Dataset used in this study was from PhysioNet, obtained from long-term EEG and ECG record of epileptic patient without history of cardiac disease. With the ability to do the identification of epilepsy phase, it is expected to help doctors and medical staffs to differ epileptic ECG signals for every different phases in epilepsy, and to prove the hypothesis whether the three phases in epilepsy can be distinguished from the heart signal.

In medical Science field, the EEG signal classification is present with a plethora of applications. The health monitoring is depending on modern technology like EEG and ECG signal recording. The EEG signals are analyzed to identify the abnormal condition of the human brains. The unusual activity of the brain is known as the seizure The electrical signal generated in the braincauses epilepsy. In this proposed work, a deep learning model Bi-LSTM is projected for the epilepsy signal classification. The Bonn university EEG dataset is used for the testing purpose. The EEG signal classification has three significant steps data pre-processing, features extraction, and classification. Data pre-processing is done by DCT and filter conversion. The Hurst exponent and ARMA feature sets are extracted from the pre-process EEG signal. A Bayesian optimization tuned Bi-LSTM model is suggested for the EEG signal classification task. The epileptic EEG signals are recognized by the proposed method. The hyperparameters of the Bi- LSTM model is tuned by the Bayesian optimization rule. Three different class ictal, pre-ictal, and inter-ictal are classified from the EEG signal data. A comparative study is also provided for the epilepsy signal classification task. The classification accuracy of for ictal is 94%, pre-ictal is 92%, and inter-ictal is 91%, which more significant than the LSTM and SVM based classifier model.

Epilepsy is generally considered as a collection of neurological disorders. Electroencephalography (EEG), a useful measure for analyzing the brain's electrical activity, has been widely used for the diagnosis of epileptic seizures. In an EEG diagnosis report, synchrony of epileptiform discharges should be included in the description. Correlation coefficient analysis could be used to measure the synchrony feature. However, the existing correlation coefficient overlooks the importance of epileptiform discharges in EEG data of epilepsy. In this paper, in order to tackle this problem, we propose a novel correlation coefficient to measure the synchrony feature. This method combines the attention mechanism with the correlation coefficient. We use the first-order difference to assign the attention weights and apply the weight to the Pearson's correlation coefficient. The first-order difference can highlight the high-frequency and high-amplitude waveforms in the original time series. Therefore, epileptiform discharges could play a more important role in the calculation of the correlation coefficient. We collected the EEG of epileptic patients during the interictal period and labeled the epileptiform discharge segments for experimental tests. In our case study, the Pearson's correlation coefficient with attention weights gave better results than the direct use of Pearson's correlation coefficient.

A novel algorithm for seizure prediction based on patient specific manner was proposed to improve the accuracy of epilepsy prediction. Time-frequency features and spatial features were extracted from each channel by 4s windows with 2s overlap. A continuous 10-min sample was selected from 1h before seizure onset by preictal period selection, which achieved maximum linear separability compared with inter ictal period. The effective features selected by elastic net and effective channels selected adaptively in greedy manner were input into SVM. The algorithm is tested on MIT scalp EEG database and the database collected in Xuanwu Hospital Capital Medical University. The algorithm can achieve a sensitivity of 94.61% and a false positive rate of 0.1484/h in MIT database, and a sensitivity of 95.14% and a false positive rate of 0.1312/h in Xuanwu Hospital database. The results show that the algorithm in this paper has higher sensitivity and lower false positive rate.

Electroencephalogram (EEG) signal compression is an essential process to speed-up the medical signal transmission with reduced storage requirements, costs, and required bandwidth. The main objective of the present work is to compress the EEG signals and study the effect of the compression process on the epileptic seizure prediction. A lossy EEG data compression scheme using Discrete Cosine Transform (DCT) is applied, followed by seizure prediction. The used dataset includes healthy, pre-ictal, and ictal signals with multiple channels. The EEG signals are segmented to segments of 10 sec length. Also, the probability density functions (PDFs) for seizure prediction are measured, including amplitude, derivative, local media, local variance, and local mean PDFs. During the testing phase, only the selected bins of PDFs are used in the prediction process to identify each signal segment as pre-ictal or normal. A method of equal benefit decision fusion is carried out in the final prediction stage leading to a single sequence of decisions representing the activities of all segments. Relative to a patient-specific estimation level, this series after being filtered with a moving average filter is compared.

The Electroencephalogram recording system plays a very important role to identify seizure occurrence and its severity range. Thus, it requires long-term video monitoring and visual interpretation from expert. In this paper, an intelligent seizure prediction algorithm is proposed allowing the prediction and localization of ictal excessive discharges and the control of consciousness impairment. The Extended Kalman Filter is employed to predict seizure occurrence. Secondly, a Spherical Spline interpolation algorithm is used to enhance the EEG spatial resolution. The algorithm was tested on 24 epileptic patients from two different databases. The proposed Method predict occurrence of all seizures before an average of 8 minutes, correctly localize ictal epileptic discharges and control their propagation on the scalp. These findings highlight the capacity to automatically extract useful spatio-temporal features from Cranial EEG.

The aim of this study is to develop an efficient, reliable and automatic epileptic seizure prediction system using scalp EEG measurements with optimal feature and minimum channels. EEG data in interictal and preictal periods from the CHB-MIT dataset are used for seizure prediction. First, the original signals are decomposed into several frequency bands using a digital wavelet transform (DWT). Then, features including standard deviation (S), log of amplitude (L), quartile (Q) and coefficient of variation (CV ) are extracted. Finally, different combinations of feature vectors are fed into classifiers (support vector machine (SVM) and extreme learning machine (ELM)) to classify the above two states (preictal and interictal states). Performance analysis shows that the optimal feature is CV, the optimal sub-band is 16-31 Hz and the optimal EEG channel can be chosen as FP1-F7, T7-P7, FP1-F3, P3-O1 or P7-T7. By comparing the classification results, ELM provides a more robust and higher overall accuracy than SVM, and the best average accuracy of both ELM and SVM can reach as high as 100%.

in this paper, the electroencephalogram (EEG) signal that consists of the information about neurological activities of the brain has been considered to detect epileptic seizures. In connection with neural electrical activities, there is a very limited study using graph theory for seizure detection. In this work, a complete regular weighted undirected graph has been constructed in which the nodes are selected based on the variance among different channels of EEG signals, whereas weights of the edges are assigned based on the phase synchronization between associated nodes. Further, several properties of the graph have been extracted and used as feature sets to perform classification tasks. The well-known classifier k-NN has been applied for classification. In addition, principal component analysis has been used to reduce the dimension of feature sets that leads to improving the analysis and classification performance. The experimental analysis has been performed using EEG signals from the PhysioNet database for three cases — pre-ictal, ictal, and post-ictal. The results demonstrate that the proposed method achieved classification accuracy by up to 100%. The other two parameters' sensitivity and specificity are approximately 100%. The results demonstrate the usability and reliability of the proposed idea.

This paper introduces a new seizure detection method based on EEG signals using Hilbert Transform (HT) and Kolmogorov Complexity (KC). The new method is named Blending Hilbert–Kolmogorov (Blending HT-KC) approach. In the Blending HT-KC method, the EEG signal is converted firstly into a complex helical sequence and then into a binary sequence. From the analytic signal, the instantaneous amplitude-based features are obtained and, from the logical signal, the KC features are extracted. The extracted attributes will then fed into the Support Vector Machine Classifier (SVM). The proposed approach is examined for four classification problems to distinguish between Ictal (during seizure) and normal states as well as between ictal and interictal (between seizures) states of epilepsy. The experimental results demonstrate the excellent performance and suitability of the proposed methodology for the detection of seizure activity.

The purpose of this study is to analyse the ictal variations in peripheral blood flow using photoplethysmogram (PPG) and single lead Electrocardiogram (ECG) signals. 11 subjects with 56 partial seizures were recorded with the PPG sensor worn on their left ankles. 6 different features from PPG pulse morphology related to hemodynamics were derived. The seizures were divided into two groups based on the side of the seizure activity. The investigation of ictal variations in features did not show any significant difference between the seizures' lateralizations. The analysis of latencies of ictal changes in the PPG features revealed the PPG pulse amplitude precede the variations in other PPG features including ictal heart rate variability. In addition, analysis of the effect of seizure lengths on ictal variations showed the seizures' lengths have no significant effect on the feature variation rates.Clinical relevance- Analysis of the extracted PPG features and their timing suggest an increase in vascular resistance due to increase in sympathetic tone which occurs prior to the ictal tachycardia. These variations is independent of the seizures' lengths and lateralizations.

This paper aims at analysing the electroencephalogram of the local, epileptic population of Chennai. The subjects used for this study suffer from different types of epilepsy. Their EEG was recorded at a sampling rate of 256 Hz. 10 second durations of these signals were exported and preprocessed using a moving average filter to remove the noise, and filtered with a band pass filter to pass the required frequency range. Wavelet denoising using Daubechies wavelet removed the artifacts in the signal, and the denoised signal was used for feature extraction. The features used are the Hjorth parameters, which characterize the power and frequency of the signal. These parameters were estimated for pre-interictal, interictal and post-interictal states. The correlation between the Hjorth parameters were plotted for all the patients, electrode wise, in order to understand the dominant electrode for the population sampled, in case of an epileptic event. The correlation proved that general epilepsy was the most common type of epilepsy existing in the sample population. It was inferred after calculating the p-values, that the Hjorth parameters characterize the epileptic signal excellently.

The absence of epileptiform activity in a scalp electroencephalogram (EEG) recorded from a potential epilepsy patient can cause delays in clinical care delivery. Here we present a machine-learning-based approach to find evidence for epilepsy in scalp EEGs that do not contain any epileptiform activity, according to expert visual review (i.e., "normal" EEGs). We found that deviations in the EEG features representing brain health, such as the alpha rhythm, can indicate the potential for epilepsy and help lateralize seizure focus, even when commonly recognized epileptiform features are absent. Hence, we developed a machine-learning-based approach that utilizes alpha-rhythm-related features to classify 1) whether an EEG was recorded from an epilepsy patient, and 2) if so, the seizure-generating side of the patient's brain. We evaluated our approach using "normal" scalp EEGs of 48 patients with drug-resistant focal epilepsy and 144 healthy individuals, and a naive Bayes classifier achieved area under ROC curve (AUC) values of 0.81 and 0.72 for the two classification tasks, respectively. These findings suggest that our methodology is useful in the absence of interictal epileptiform activity and can enhance the probability of diagnosing epilepsy at the earliest possible time.

Epilepsy diagnosis through visual examination of interictal epileptiform discharges (IEDs) in scalp electroencephalogram (EEG) signals is a challenging problem. Deep learning methods can be an automated way to perform this task. In this work, we present a new approach based on convolutional neural network (CNN) to detect IEDs from EEGs automatically. The input to CNN is a combination of raw EEG and frequency sub-bands, namely delta, theta, alpha and, beta arranged as a vector for one-dimensional (1D) CNN or matrix for two-dimensional (2D) CNN. The proposed method is evaluated on 554 scalp EEGs. The database consists of 18,164 IEDs marked by two neurologists. Five-fold cross-validation was performed to assess the IED detectors. The resulting 1D CNN based IED detector with multiple sub-bands achieved a false positive rate per minute of 0.23 and a precision of 0.79 at 90% sensitivity. Further, the proposed system is evaluated on datasets from three other clinics, and the features extracted from CNN outputs could significantly discriminate (p-values < 0.05) the EEGs with and without IEDs. We have proposed an optimized method with better performance than the literature that could aid clinicians to diagnose epilepsy expeditiously, and thereby devise proper treatment.

Epilepsy is a chronic neurological disorder where seizures occur due to the transient and unexpected electrical disturbances of the brain. Epilepsy affects the quality of life of patients, making them unable to carry out essential day to day activities. Predicting an impending seizure will help them to make effective interventions. In this work, we proposed an automated system to classify the EEG data into ictal, nonictal, and pre-ictal classes using ResNET-50, a subclass of convolutional neural networks, by converting 1D EEG data into 2D EEG images. Using this novel approach, the proposed model predicts an oncoming seizure with an accuracy of 94.98%. This shows that deep residual network is a promising approach for analyzing EEG data to predict epileptic seizures.

Seizure prediction can allow timely preventive measures for patients with epilepsy. In this study, we propose a hybrid model consisting of convolutional neural networks (CNNs) and an extreme learning machine (ELM) to predict seizures using scalp EEG. We first covert the EEG time series on 30-s windows into 2D spectrograms using the short-time Fourier transform. Then we apply CNNs to these images to extract features automatically. Finally, we use the ELM to classify preictal and interictal segments. The proposed method achieves sensitivity of 95.85% and a false prediction rate of 0.045/h on the Boston Children’s Hospital-MIT scalp EEG dataset.

Epilepsy is a critical symptomatic attribute of cerebral dysfunction where the performance of nervous cells gets disturbed. The basic character of seizure or epilepsy is unforeseen and results in a misunderstanding by the patient and onlookers in the family or elsewhere causing insecurity and unsafe conditions for all concerned. This causes damage and injury to major organs, hence a life threat. Prediction and detection of seizures is a vital aspect. Its prediction and warning could help avoid any type of epileptic stages to protect the patient from internal losses owing to such attacks is a challenge for the medical and surgical fields in minimizing risk management to enhance a better lifestyle. This work ensures the causes, prediction and detection to find the reasons behind epileptic conditions and necessarily elaborates conditions of seizures, its stages and types viz. preictal, ictal, interictal, postictal respectively.

Epilepsy affects over 50 million people worldwide and 30% of patients' seizures are medically refractory. The process of localizing and removing the epileptogenic zone is error-prone and ill-posed in part because we do not understand how epilepsy manifests. It has recently been proposed that the epileptic cortex is fragile in the sense that seizures manifest through small perturbations in the synaptic connections that render the entire cortical network unstable. If the fragility of the cortical network could be computed over a period in which seizure genesis occurs, then it might elucidate network mechanisms correlated to the epileptogenic zone. In this study, we used local field potentials (LFP) from neocortex by implementing an acute model of epilepsy in mice. These recordings were used to develop a dynamical network model that quantifies the fragility of the nodes from LFP epochs of baseline activity, preictal and ictal states. Fragility was quantified by the generation of a linear time-varying model to which we then applied a perturbation to determine the sensitivity of nodes in the network. Spatiotemporal fragility maps showed clear quantifiable changes in the epileptogenic network's properties throughout different states of seizure genesis. We quantified this difference over a baseline, preictal and ictal periods to show that network fragility is modulated in the manifestation of epilepsy.

Epilepsy is the most rampant neurological disorder that is diagnosed for around 50 million people in the world. We obtain beneficial information related to the different physiological states of the brain via the electroencephalogram (EEG). S pikes in an EEG recording serves as a major parameter for diagnosing epilepsy in an individual [2]. This research aims to efficiently identify and classify Spikes in an EEG Signal using multilayer filtration of noises to predict epilepsy. The methodology explained in the following research introduces two layers of a classifier to classify time series into spikes and nonspikes given a channel of EEG recording. The advantages of the proposed method are its computational simplicity, easy algorithmic implementation, and immunity to artifacts, noise, and implicit characteristics of EEG. As opposed to a naive approach of sliding windows for the classification of time series, the following method uses a layer of the preliminary classifier with very high classification sensitivity and a layer of classification, where the focus is given on high classification accuracy. The detection algorithm uses mathematical operations and thresholding in the first layer and uses the support vector machine(SVM) classifier in the second layer to classify the time series windows into spikes and non-spikes in an EEG recording.

Today, the security of wearable and mobile-health technologies represents one of the main challenges in the Internet of Things (IoT) era. Adversarial manipulation of sensitive health-related information, e.g., if such information is used for prescribing medicine, may have irreversible consequences involving patients’ lives. In this article, we demonstrate the power of such adversarial attacks based on a real-world epileptic seizure detection problem. We identify the minimum perturbation required by the adversaries to declare a seizure (ictal) sample as non-seizure (inter-ictal) in emergency situations, i.e., minimal adversarial perturbation to fool the classification algorithm.

Epilepsy is a disorder in the electrical activity of the brain that occurs in a specific area or even the entire brain. These changes are visible through the acquisition of electroencephalogram (EEG) brain signals. EEG signals are important tools in predicting epilepsy because they are noninvasive measurement and display electrical activity at different external nodes at human brain. We used the CHB-MIT EEG Database in this study to develop an artificial model to predict epileptic seizures. Thus, we applied a one-dimensional convolutional neural network (CNN) to investigate raw EEG signals as an important indicator for starting time of a seizure. The seven-layer CNN was used to detect Preictal and Interictal states of brain where the performance of the proposed model was evaluated in terms of accuracy, specificity, and sensitivity which resulted in 97%, 98.47%, and 98.5%, respectively. Moreover, the proposed model was trained in two different feeding states: 1-Feeding by individual channel, 2-Feeding by grouped channels. It seems that the obtained results are promising.

This work presents beta subband (12-30 Hz) as a biomarker to distinguish between interictal and ictal states using Haralick features. Previous works has showed whole frequency spectrum for this analysis. Significance of this work is it has used only beta subband of electroencephalogram (EEG) for classification using image descriptors. One dimensional EEG data has been converted into image using Short-time Fourier Transform (STFT). Beta subband is cut from the time frequency (t-f) plane and Haralick features is fed in the decision tree classifier. The results have been evaluated using K-fold cross validation and classification accuracy of 92.5% has been calculated. Receiver operating characteristic (ROC) analysis has also been performed which shows maximum area under curve (AUC) of 0.94 to distinguish between interictal and ictal.

Epilepsy is a neurological disorder that causes abnormal behavior and recurrent seizures due to unusual brain activity. This study has attempted to predict seizures in epileptic patients through the process of feature extraction from EEG signals during preictal/ictal and interictal periods, classification and regularization. EEG signals from various parts of the brain from 10 epileptic patients are considered. Fast Fourier Transform (FFT) is used to determine the three features-the phase angle, the amplitude and the power spectral density of the signals. To classify the signals, these features are then used along with Support Vector Machine (SVM) as the classifier. Furthermore, regularization is used to make better predictions i.e. increase prediction accuracy and decrease the rate of false alarm. Finally, the proposed approach is tested on CHB-MIT Scalp EEG data set and it is able to predict epileptic seizures 25 minutes on average before the onset of the seizure with 100% accuracy and a low false-alarm rate of 0.46 per hour. This study intends to contribute to the development of better and advanced seizure predicting devices in the medical field.

Epilepsy is a fatal disease worldwide, which is considered as the second most common brain disease. It can be represented by an electroencephalogram (EEG) signal for epileptic seizure analysis. In this work, features are extracted from S-transform of EEG signals with fixed periods of time. These features are extracted from three states, namely normal, pre-ictal and ictal (seizure). The powers of the different EEG extracted features are calculated. Afterwards, the support vector machines (SVMs) are applied to distinguish between the different periods of the different states. The simulation results reveal the impact of using S-transform for seizure detection with average sensitivity and specificity of 94.481%, and 70.315%, respectively. Moreover, the seizure prediction is performed accurately with average sensitivity and specificity of 96%, and 72.944%, respectively.

Epilepsy is a type of neurological disorder which can happen without serious warning and affects people almost at any age. It is a brain disorder caused by sudden and unprovoked seizures as a result of excitation of a lot of brain cells simultaneously which may lead to physical symptoms abnormalities and deformation such as failure in concentration, memory, attention etc. therefore, proper and efficient method of continues monitoring and detection of these epileptic seizures is paramount. This work presents an effective and efficient technique suitable for smart, low cost, power and real time devices that can be easily integrated with recent 5G network IoT devices for mobile applications, home and health care centers for monitoring and alert the doctors and patients about its occurrence to prevent a sudden collapse and consciousness which may cause injury and death. We proposed a low computational cost features extraction method by utilizing the efficacy of time-frequency, statistical and non-linear features known as hybrid techniques. The efficiency and accuracy of these smart devices is highly depends on quality of feature extraction methods and classifier performance. Therefore, this work employed two machine learning classifiers, support vector machine (SVM) and feedforward neural network (FFNN) to detect and classify interictal (normal) and ictal (seizure) signals. Discrete wavelet transform (DWT) is employed to decomposes the signals into decomposition levels as sub-bands of the signals to capture the non-stationarity of the EEG signals. Mean, median, maximum, minimum etc. were calculated for each sub-band as statistical parameters, non-linear features such as sample entropy, approximate entropy and wavelet energy were also calculated. The combination of features is then fed to two classifiers for the classification. Based on the performance measures such as accuracy, sensitivity and specificity, our proposed approach reveals a promising result with highest accuracy of 99.6%.