epilepsy seizure disorder research paper

• Open access
• Published: 25 May 2020

A review of epileptic seizure detection using machine learning classifiers

• Mohammad Khubeb Siddiqui   ORCID: orcid.org/0000-0001-6699-6216 1   na1 ,
• Ruben Morales-Menendez 1 ,
• Xiaodi Huang 2   na1 &
• Nasir Hussain 3  

Brain Informatics volume  7 , Article number:  5 ( 2020 ) Cite this article

21k Accesses

204 Citations

1 Altmetric

Metrics details

Epilepsy is a serious chronic neurological disorder, can be detected by analyzing the brain signals produced by brain neurons. Neurons are connected to each other in a complex way to communicate with human organs and generate signals. The monitoring of these brain signals is commonly done using Electroencephalogram (EEG) and Electrocorticography (ECoG) media. These signals are complex, noisy, non-linear, non-stationary and produce a high volume of data. Hence, the detection of seizures and discovery of the brain-related knowledge is a challenging task. Machine learning classifiers are able to classify EEG data and detect seizures along with revealing relevant sensible patterns without compromising performance. As such, various researchers have developed number of approaches to seizure detection using machine learning classifiers and statistical features. The main challenges are selecting appropriate classifiers and features. The aim of this paper is to present an overview of the wide varieties of these techniques over the last few years based on the taxonomy of statistical features and machine learning classifiers—‘black-box’ and ‘non-black-box’. The presented state-of-the-art methods and ideas will give a detailed understanding about seizure detection and classification, and research directions in the future.

1 Introduction

The word epilepsy originates from the Latin and Greek word ‘epilepsia’ which means ‘seizure’ or ‘to seize upon’. It is a serious neurological disorder with unique characteristics, tending of recurrent seizures [ 1 ]. The context of epilepsy, found in the Babylonian text on medicine, was written over 3000 years ago [ 2 , 3 ]. This disease is not limited to human beings, but extends to cover all species of mammals such as dogs, cats and rats. However, the word epilepsy does not give any types of clues about the cause or severity of the seizures; it is unremarkable and uniformly distributed around the world [ 1 , 4 ].

Several theories about the cause are already available. The main cause is electrical activity disturbance inside a brain [ 1 , 5 , 6 ], which could be originated by several reasons [ 7 ] such as malformations, shortage of oxygen during childbirth, and low sugar level in blood [ 8 , 9 ]. Globally, epilepsy affects approximately 50 million people, with 100 million being affected at least once in their lifetime [ 5 , 10 ]. Overall, it accounts for 1% of the world’s burden of diseases, and the prevalence rate is reported at 0.5–1% [ 4 , 11 ]. The main symptom of epilepsy is to experience more than one seizure by a patient. It causes a sudden breakdown or unusual activity in the brain that impulses an involuntary alteration in a patient’s behaviour, sensation, and loss of momentary consciousness. Typically, seizures last from seconds to a few minute(s), and can happen at any time without any aura. This leads to serious injuries including fractures, burns, and sometimes death [ 12 ].

1.1 Seizure type

Based on the symptoms, seizures are categorized by neuro-experts into two main categories—partial and generalized [ 7 , 13 ]—as shown in Fig.  1 . Partial seizure, also called ‘focal seizure’, causes only a section of the cerebral hemisphere to be affected. There are two types of Partial seizure: simple-partial and complex-partial. In the simple-partial, a patient does not lose consciousness but cannot communicate properly. In the complex-partial, a person gets confused about the surroundings and starts behaving abnormally like chewing and mumbling; this is known as ‘focal impaired awareness seizure’. On the contrary, in the generalized seizures, all regions of the brain suffer and entire brain networks get affected quickly [ 14 ]. Generalized seizures are of many types, but they are broadly divided into two categories: convulsive and non-convulsive.

Types of seizure. Showing types of seizure and its sub-types

1.2 Main contributions of the paper

In brief, the contributions of this paper are as follows:

We have done the review according to five main dimensions. First, researchers who adopted the EEG, ECoG or both for seizure detection; second, significant features; third, machine learning classifiers; fourth, the performance of the classifier during a seizure, and last, knowledge discovery (e.g., seizure localization).

Through study, it has been explored that an ensemble of decision trees (i.e., decision forest–random forest) classifier outperforms other classifiers (ANN, KNN, SVM, single Decision Tree).

We also suggest, how decision forest algorithms could be more effective for other knowledge discovery tasks besides seizure detection.

This study will help the researchers with their data science backgrounds to identify which statistical and machine learning classifiers are more relevant for further improvement to the existing methods for seizure detection.

The study will also help the readers for understanding about the publicly available epilepsy datasets.

In the end, we have provided our observations by the current review and suggestions for future research in this area.

The structure of the paper is organized as follows. “ Role of data scientists in epileptic seizure detection ” section gives the overview of machine learning experts in EEG datasets. The preliminaries requirements are provided in “ A framework for seizure detection ” section; it presents a general model of seizure detection and explains each step in a subsequent manner. “ Publicly available datasets ” section provides the details of benchmark datasets with their description. “ Seizure detection based on statistical features and machine learning classifiers ” section explains the review of literature work done on seizure detection using different machine learning classifiers, with a detailed comparison. “ Seizure localization ” section reviews the work done in identifying the affected lobes of the brain using machine learning classifiers. In “ Problems identified in existing literature ” section, we have explored the issues in the previous work and highlighted the gap. Overall, “ observation about capable classifiers and statistical features ” section reports our observations from the review about a suitable classifier and feature. “ Research directions in seizure detection ” section emphasizes the future directions in this research area, followed by “ Conclusion ” section on the summary of the paper.

2 Role of data scientists in epileptic seizure detection

Applications of machine learning are significantly seen on health and biological data sets for better outcomes [ 15 , 16 ]. Researchers/scientists on different areas, specifically, data mining and machine learning, are actively involved in proposing solutions for better seizure detection. Machine learning has been significantly applied to discover sensible and meaningful patterns from different domain datasets [ 17 , 18 ]. It plays a significant and potential role in solving the problems of various disciplines like healthcare [ 17 , 19 , 20 , 21 , 22 , 23 , 24 , 25 ]. Applications of machine learning can also be seen on brain datasets for seizure detection, epilepsy lateralization, differentiating seizure sates, and localization [ 26 , 27 , 28 , 29 ]. This has been done by various machine learning classifiers such as ANN, SVM, decision tree, decision forest, and random forest [ 26 , 28 ].

Certainly, in the past, numerous reviews have been carried out on seizure detection along with applied features, classifiers, and claimed accuracy [ 27 , 30 , 31 , 32 , 33 ] without focusing on the challenges faced by the data scientists whilst doing research on datasets of neurological disorders. Therefore, this article provides a detailed study of machine learning applications on epileptic seizure detection and other related knowledge discovery tasks. In this review, the collected articles are from well-known journals of their relevant field. These references are either indexed by SCOPUS or Web of Science (WOS) . Besides, we also considered some good ranked conference papers. Extensive literature is available covering the deep analysis of different features and classifiers applied on EEG datasets for seizure detection [ 31 , 34 , 35 ]. Both, feature extraction and applying classification techniques are challenging tasks. Previous literature reveals that for the past few years, interest has been increased in the application of machine learning classifiers for extracting meaningful patterns from EEG signals, which helps for detecting seizures, its location in the brain, and other impressive related knowledge discoveries [ 28 , 36 , 37 ]. Three decades ago, Jean Gotman [ 6 , 38 , 39 , 40 ], analyzed and proposed the model for effective usage of EEG signals by applying different computational and statistical techniques for automatic seizure detection. Furthermore, the research has been carried out by different signal processing methods and data science methods to provide better outcomes [ 27 , 34 , 41 , 42 , 43 , 44 , 45 , 46 , 47 ].

3 A framework for seizure detection

In this section, we present a pictorial framework of the model used for seizure detection from an EEG/ECoG seizure dataset, illustrated in Fig.  2 . The process comprises four steps: Data Collection, Data Preparation, Applying Machine Learning Classifiers and Performance Evaluation.

Basic model of epileptic seizure detection. This explains the basic steps to collect the dataset by EEG medium, display of raw EEG signals, transform EEG signals to two-dimensional table, feature selection, prepare the dataset with seizure (S) and non-seizure (NS) , apply machine learning classifier(s) and seizure detection, or other related tasks

3.1 Data collection

The initial requirement is to collect the dataset of brain signals. For this, different monitoring tools are used. Typically, the mostly used devices are EEG and ECoG, because their channels or electrodes are implanted by glue on the surface of the scalp as per 10–20 International system [ 48 ] at different lobes. Each of them has a wire connection to the EEG device, providing timely information about the variations in voltage, along with temporal and spatial information [ 49 ]. As highlighted in Fig.  2 , the EEG channels are placed on the subject’s scalp, and the electrical signals are read by the EEG monitoring tool and it displays these raw signals over the screen. Further, these raw signals have been carefully monitored by the analyst and classified into ‘seizure’ and ‘non-seizure’ states.

3.2 Data transformation

After data collection, the next crucial step is to transform the signal data into a 2-D Table format. The reason for this is to make it easier for analysis and provide necessary knowledge like seizure detection. This datum is raw because it has not been processed yet. Therefore, it will not be suitable to give relevant information. To do the processing, different feature selection modalities have been applied. This step also presents the dataset as supervised, which means that it provides the class attribute with possible class-values.

3.3 Dataset preparation

For data transformation, data processing is a decisive step to extract meaningful information from the collected raw dataset. As such, different feature extraction techniques have been used; as shown in Table  1 . These methods are generally applied to the extracted EEG signal dataset [ 31 , 34 ]. The raw dataset becomes rich in terms of different statistical measure values.

After feature extraction processing, the dataset becomes more informative that it ultimately helps the classifier for retrieving better knowledge.

3.4 Applying machine learning classifiers and performance evaluation

To achieve a high accuracy of seizure detection rate and explore relevant knowledge from the EEG processed dataset, different supervised and unsupervised machine learning have been used.

3.4.1 Classification

In classification, a dataset D has a set of ‘non-class attributes’, and a ‘class attribute’. They are the principal components and their pertinent knowledge is very important, as both have a strong association for potential classification. The target attribute is defined as the ‘class attribute’ C , and it comprises more than one class values, e.g., seizure and non-seizure . On the contrary, attributes \(A=\{A_1,A_2.A_3 \ldots A_n\}\) are known as ‘non-class attributes’ or predictors [ 50 , 51 ]. The following classifiers have been popularly used in seizure detection. Common classifiers such as SVM [ 52 ], decision tree [ 53 ] and decision forest [ 54 ] are applied to the processed EEG dataset for seizure detection.

3.4.2 Performance evaluation

The accuracy of the obtained results is used to evaluate different methods. The most popular training approach is tenfold cross-validation [ 55 ], where each fold, i.e., one horizontal segment of the dataset is considered to be the testing dataset and the remaining nine segments are used as the training dataset [ 56 , 57 ].

Except for the accuracy, the performance of the classifiers is commonly measured by the following metrics such as precision, recall, and f-measure [ 58 ]. These are based on four possible classification outcomes—True-Positive (TP), True-Negative (TN), False-Positive (FP), and False-Negative (FN) as presented in Table  2 .

Precision is the ratio of true-positives to the total number of cases that are detected as positive (TP+FP). It is the percentage of selected cases that are correct, as shown in Eq.  1 . High precision means the low false-positive rate.

Recall is the ratio of true-positive cases to the cases that are actually positive. Equation  2 shows the percentage of corrected cases that are selected.

Despite getting the high Recall results of the classifier, it does not indicate that the classifier performs well in terms of precision. As a result, it is mandatory to calculate the weighted harmonic mean of Precision and Recall; this measure is known as F-measure score, shown in Eq.  3 . The false-positives and the false-negatives are taken into account. Generally, it is more useful than accuracy, especially when the dataset is imbalanced.

4 Publicly available datasets

For data scientists and researchers, a dataset used is important for evaluating the performance of their proposed models. Similarly, in epileptic seizure detection, we need to capture the brain signals. EEG recording is the most used method for monitoring brain activity. These recordings play a vital role in machine learning classifiers to explore the novel methods for seizure detection in different ways such as onset seizure detection, quick seizure detection, patient seizure detection, and seizure localization. The significance of publicly available datasets is that they provide a benchmark to analyze and compare the results to others. In the following section, we will describe the popular datasets that are widely used on epilepsy.

4.1 Children Hospital Boston, Massachusetts Institute of Technology—EEG dataset

This dataset is publicly available on a physionet server and prepared at Children Hospital Boston, Massachusetts Institute of Technology (CHB-MIT) [ 59 , 60 ]. It can be collected easily via Cygwin tool which interacts with the physionet server. It contains the number of seizure and non-seizure EEG recordings for each patient of the CHB [ 61 ]. The dataset comprises 23 patients; 5 males, aged 3–22 years, and 17 females aged 1.5–19. Each patient contains multiple seizure and non-seizure recording files in European data format (.edf), representing the spikes with seizure start and end time, which is easily visible at a browser called an ‘EDFbrowser’. The primary datasets are in the 1-D format, containing EEG signals that are obtained through the different types of channels that were placed on the surface of the brain as per 10-20 International System. All these signals of the dataset were sampled at the frequency of 256Hz.

4.2 ECoG Dataset, Epilepsy Centre, University of California

This is a publicly available dataset of electrocorticogram (ECoG) signals from an epileptic patient, which was collected from the Epilepsy Center, University of California, San Francisco (UCSF) [ 62 ]. It was originally collected by implanting 76 electrodes on the scalp in both invasive (12-electrodes) and non-invasive manner (64-electrodes). It comprises 16 files altogether. Out of these, eight files ( \(F1, F2, \cdots F8\) ) are classified as ‘pre-ictal’ meaning the stage before the seizure. The rest of the files ( \(F9, F10, F11, \cdots F16\) ) represent the ‘ictal’ stage data. The collected data are sampled at the frequency of 400 Hz (i.e., 400 cycles/s) and the total duration is 10 s. As a result, there are (400 cycles/s \(\times\) 10 s) 4000 cycles in each file [ 63 ].

4.3 The Freiburg—EEG dataset

This dataset was collected from the invasive EEG recordings of 21 patients (8 males aged 13–47 years, 13 females aged 10–50 years) suffering from medically intractable focal epilepsy. It was recorded during an invasive pre-surgical epilepsy monitoring at the Epilepsy Centre of the University Hospital of Freiburg, Germany [ 64 ]. Out of 21 patients, 13 patients had 24 h of recordings, and 8 patients had less than 24 h. These recordings are inter-ictal, and together they provide 88 seizures.

4.4 Bonn University—EEG dataset

The dataset comprises five subsets, where each one denoted as (A–E) contains 100 single-channels recording, and each of them has a 23.6 s duration, captured by the international 10–20 electrode placement scheme. All the signals are recorded with the same 128-channel amplifier system channel [ 65 ].

4.5 BERN-BARCELONA—EEG dataset

This dataset comprised EEG recordings derived from five pharmacoresistant temporal lobe epilepsy patients with 3750 focal and 3750 non-focal bivariate EEG files. Three patients were seizure-free, with two patients only having auras but no other seizures following surgery. The multichannel EEG signals were recorded with an intracranial strip and depth electrodes. The 10–20 positioning was used for the electrodes’ implantation. EEG signals were either sampled at 512 or 1024 Hz, depending on whether they were recorded with more or less than 64 channels. According to the intracranial EEG recordings, they were able to localize the brain areas where seizures started for all five patients [ 66 ]. This dataset is good for the seizure localization purpose.

5 Seizure detection based on statistical features and machine learning classifiers

This section explains the comprehensive detail of work on seizure detection using statistical features, classifiers—‘black-box’ and ‘non-black-box’. They are illustrated in Table  3 . In brief, the ‘black-box’ classifiers are those which provide the accuracy without mentioning the reasons behind the results such as ANN and SVM [ 67 ]. They are unable to explain their classification steps. Whereas, ‘non-black-box’ classifiers such as decision forest and random forest can able to explain each step of the processing, which is human-understandable. As a result, it helps in human-interpretable knowledge with high accuracy [ 68 ].

5.1 Seizure detection based on statistical features

If we apply machine learning classifier(s) directly to raw EEG/ECoG datasets, it may not produce enough sensible patterns. Therefore, selecting significant and capable statistical features from EEG and ECoG raw datasets is one of the challenges and a crucial task. The nature of EEG and ECoG signals is very complex, non-stationary and time-dependent [ 105 , 106 , 107 ]. As such, we can apply the machine learning classifier(s) to the processed datasets, which will ultimately assist to solve various neurological problems; for example, identifying seizure’s stages, accurate seizure detection, fast detection, etc. In Table  3 , we summarize a review of several studies.

The significant statistical features were extracted by different types of transformation techniques; discrete wavelet transformations (DWT), continuous wavelet transformation (CWT), Fourier transformation (FT), discrete cosine transformation (DCT), singular value decomposition (SVD), intrinsic mode function (IMF), and time–frequency domain from EEG datasets [ 34 , 71 , 79 , 108 ]. Logesparan et al. [ 34 ] used different types of feature extraction methods for seizure detection, but they reported that two features—‘line length’ and ‘relative power’—are the good performers for seizure detection. Guerrero-Mosquera [ 109 ] applied three time-domain features—line length, frequency, and energy on the raw EEG dataset. These features claim to be suitable for seizure detection and other brain-related applications such as computer interface (BCI). The claimed performance was evaluated using the following metrics such as sensitivity, specificity, F-score, receiver operating characteristics (ROC) curve, and percentile bootstrap measures. Duo Chen [ 84 ] used DWT with the SVM classifier on two benchmark datasets—CHB-MIT and Bonn University, achieved seizure detection accuracies of 92.30% and 99.33%, respectively. Ramy Hussein et al. [ 100 ] proposed a new featured L1-penalized robust regression (L1PRR) for seizure detection, the issue with their approach is computational complexity. Zavid and Paul [ 99 ] focused on classifying the ‘ictal’ and ‘inter-ictal’ states, where they used four features DCT, DCT-DWT, SVD, and IMF; the obtained signals are further classified by LS-SVM due to less computational cost.

Several researchers have contributed to seizure detection using a single feature [ 108 , 110 ]. The feature ‘line length’ [ 108 , 110 ] was applied to an EEG dataset; approximately 4.1 s of mean detection latency is recorded at a false alarm rate of 0.051 Fp/h. Further, Guo et al. [ 69 ] also used ‘line length’ but with the ANN for classifying the records obtained by EEG signals. Their automated seizure detection accuracy is 99.6%. A system was proposed by Koolen et al. [ 70 ] to detect seizures from EEG recordings. This detection system uses a single feature—‘line length’. The performance of this system shows 84.27% accuracy, 84.00% sensitivity and 85.70% specificity, which are comparatively lower than the results of Guo et al. [ 69 ].

After 3 years of study on several of statistical features [ 34 ], Logesparan et al. [ 71 ] proposed the ‘line length’ feature for normalization and discrimination of class values from EEG datasets. It is noted that ‘line length’ could be taken as the strongest feature and provides considerable output. Based on previous studies, the ‘line length’ can be taken with other features, and the result would be more promising, specifically in machine learning. This is because the dataset dimension would also increase with meaningful statistical information in the attributes.

Some other studies on seizure detection based on a single feature, i.e., entropy and its sub-types such as approximate entropy (AE) and sample entropy (SE), have also been done [ 45 , 72 , 73 , 111 ]. The entropy feature helps to find the random behaviour of EEG signals and takes depth benefits in measuring the impurity of the signals [ 112 , 113 ]. The entropy feature has been used widely where data are in the form of signals such as ECG, [ 114 ], EEG, and ECoG [ 36 ]. This helps in further steps of the detection model.

Acharya et al. [ 111 ] used four different types of entropy-based features: sample entropy, approximate entropy, phase entropy (S1), and phase entropy (S2) of the EEG datasets. The processed dataset from these entropy features was used for seizure detection. In another study, Chen et al. [ 90 ] used eight different kinds of entropy feature—approximate, sample, spectral, fuzzy, permutation, Shannon, conditional and correction conditional on a raw EEG dataset; further, the processed data were classified into three class values: ‘ictal’, ‘inter-ictal’ and ‘normal stage’, and their accuracy is 99.50%. A tool was proposed by Selvakumari et al. [ 89 ] using four features—entropy, root mean square (RMS), variance, and energy. Based on these features, the detection was done using SVM and naïve Bayesian classifiers with a reported accuracy of 95.63%. The tool is also able to find the seizure region in the brain; however, they did not mention the exact percentage of seizure location. Song and Li [ 72 ] built classification models by two classifiers—Extreme Learner Machine (ELM) and the back-propagation neural network (BPNN). Overall, their findings show 95.6% of classification accuracy with less execution time. Yong Zhang et al. [ 73 ] applied two entropy features—AE and SE on two different classifiers—ELM and SVM for processing EEG dataset. The SE features with ELM provide good classification accuracy compared to the AE feature whilst detecting the seizure.

The energy feature has been significantly used in seizure detection [ 115 ]. It plays a vital role particularly when the seizure is detected by the epoch- or windows-based method. This means that the EEG signals are divided into various segments [ 79 , 94 ]. An exponential energy feature has been introduced by Fasil and Rajesh [ 97 ], which helps in identifying the irregularities in amplitude EEG signals.

Observations This section has provided an overview of the contributions of statistical features to seizure detection and their importance. Some researchers detect seizures using multiple sets of features, whilst others select a single feature such as ‘line length’. We recommend the ‘line length’ feature to be in the list of the set of suitable features for seizure detection because it is helpful in measuring the EEG signals complexity. It plays a sensitive role in the changes at the frequency and amplitude of signals. As a result, it helps to discriminate against the ‘seizure’ and ‘non-seizure’ cases. However, from the data science point of view, it is very important to see the various perspectives of each brain signals by observing other statistical features. Furthermore, we also suggest not to use the irrelevant feature(s) as they will unnecessarily increase the dataset size which results in an increase in computational time and gives insensible patterns too. As a result, it becomes a hassle to machine learning classifiers and users rather than providing the benefit. Some researchers [ 95 , 98 , 101 ] used a large number of features, which increases the attribute size, and results in more computational time and less accuracy. So, if we take the fewer features as previous researchers have done [ 71 , 73 , 79 ] this will give the low-dimensional dataset, which will not be fruitful for the knowledge discovery process. The next section illustrates the seizure detection by ‘black-box’ classifiers. As far as the classification purpose is concerned, it would be better to take more relevant statistical features, which can be integrated into knowledge discovery and a good performance rate.

5.2 Seizure detection based on black-box classifiers

The classifiers such as SVM, ANN, and KNN are considered as prominent ones due to their remarkable performances in different domains [ 67 , 116 ]. Each technique has its pros and cons, and ‘black-box’ methods are not an exception to this [ 104 ]. Even though these classifiers contribute well to brain datasets, some of the relevant works on seizure detection using these classifiers are reported here.

The study of Satapathy et al. [ 85 ] was based on two ‘black-box’ approaches—SVM and Neural networks using different kernel methods for seizure detection against a large EEG dataset. The performance of each classifier is measured independently by the majority voting system, and it was found that SVM was more capable than other neural networks. Subasi et al. [ 87 ] proposed the solution to detect seizure using a hybrid approach of SVM, genetic algorithm (GA), and particle swarm optimization (PSO). The method achieved impressive accuracy, i.e., 99.38%, but the problem is that the classifier trains the dataset twice, one for SVM-GA and another for SVM-PSO. This could be a time-consuming.

Shoeb and Guttag [ 41 ] performed seizure detection on their arranged dataset of Child Hospital Bostan, MIT (CHB-MIT) [ 60 ] using SVM with the vector feature and achieved the estimated accuracy of 96%. Dorai and Ponnambalam [ 42 ] came with an idea of the epoch, which means dividing the dataset into smaller time frames. Further, they applied an ensemble of four ‘black-box’ approaches—LDA, KNN, CVE, and SVM on these epoch EEG datasets. This approach provides the prediction of onset seizures 65 s earlier. Classifying the EEG data into two class ‘ ‘seizure” and ‘non-seizure’, Birjandtalab et al. [ 117 ] used a Gaussian mixture model (GMM) before detecting the seizure, and obtained 90% accuracy with 85.1% F-measure. They also raised the issue of class imbalance in their dataset. Tzallas et al. [ 103 ] used time–frequency-domain features with ANN for the EEG dataset and obtained 100% accuracy for the ‘seizure’ and ‘non-seizure’ classification problem; with epochs’ datasets the accuracy is 97.7% from (A, B, C, and D) for ‘non-seizure’ and set E for ‘seizure’ epoch classes. Amin et al. [ 79 ] extracted relative energy features from the DWT method, and four classifiers—SVM, MLP, KNN, and Naïve Bayes—were applied for the classification purpose, the result shows 98% of SVM accuracy, which outperforms remaining classifiers. A framework had been proposed by K. Abualsaud et al. [ 118 ] using the ensemble of ‘black-box’ classifiers for automated seizure detection on noisy EEG signals, and the reported classification accuracy is 95%. However, the ensemble approach did not provide good accuracy as desired because all four classifiers were ‘black-box’.

In 2018, Lahmiri et al. [ 92 ] used generalized Hurst exponent (GHE) and KNN, to propose a system for identifying the ‘seizure’ and ‘non-seizure’ classes from intracranial EEG recordings, detection rate, with 100% accuracy rate. Further, Lahmiri et al. [ 43 ] exploited GHE with SVM, to classify the ‘seizure’ and ‘non-seizure’, and also they found 100% accuracy in less time. Here, the good indication is that authors claim the good accuracy in less time for seizure detection. But, the authors did not clearly define how many times the seizure can be detected. In another study by Al Ghayab et al. [ 88 ], the obtained accuracy is 100% as a result of using the concept of Information gain theory, to extract and rank the meaningful features from EEG signal dataset. The least square-support vector machine (LS-SVM) is then applied to classify the seizure cases. Moreover, due to the ‘black-box’’s nature of applied classifiers, the authors could not explore any other related aspects in terms of Knowledge discovery. Zabihi et al. [ 81 ] did patient-specific seizure detection using SVM classifier on the processed dataset with a good set of features, comprising time-domain, frequency-domain, time–frequency domain, and non-linear feature. The performance of their model has achieved an average of 93.78% sensitivity and a specificity of 99.05%. Here, it is noteworthy that they skip an important feature—‘line length’, from the available literature, which is prominently used in seizure detection. We also argue that CHB-MIT dataset [ 60 ] is imbalanced because, in an hour(s) of recording, a seizure time span is for a few seconds.

Observations

The main issue with ‘black-box’ classifiers is that they only make prediction without providing logic rules or patterns. That is why, they are not recommended for extracting sensible knowledge. For example, for class imbalance issues in EEG datasets, insufficient related literature is found, and the researchers who attempted to work on this problem did not provide a conceivable solution as to how to solve the class imbalance issue whilst detecting the seizure.

5.3 Seizure detection based on non-black-box classifiers

‘Black-box’ classifiers are unable to express their classification procedure for human interpretation [ 67 , 104 , 116 ]. Consequently, there are fewer chances for knowledge discovery and better accuracy performance. Therefore, the concept of ‘non-black-box’ classifiers such as decision trees, and decision forests came into practice.

Chen et al. [ 119 ] first introduced the decision tree to the EEG dataset for seizure detection. Kemal and Saleh [ 120 ] used a C5.0 decision tree [ 121 ] algorithm to explore the logic rules for seizure detection, with an average accuracy of 75%. When the same C5.0 was applied to the same dataset processed by Fourier transformation the obtained accuracy with cross-validation was, however, 98.62%. A few related works are been available, where only a decision tree method is applied seizure detection because of less accuracy and a limited number of patterns obtained from the logic rules of a decision tree [ 122 ]. As a result, both the knowledge discovery and accuracy suffer. However, this gap can be filled by applying decision forest approaches instead [ 51 , 57 , 123 ].

Through the literature, it is found that the decision forest approaches are more effective than the single decision tree [ 57 , 124 ], because the decision tree often gives a confined set of rules and overfitting issue is also raised [ 68 ]. The rules are extracted from training data by a decision tree that generates either limited or a single set of logic rules (Say, wherever C2_Entropy value \(\le 101.01\) then \(Class\_value=seizure\) ) and stops growing the tree further records in the training dataset once the rule is accepted. However, if we generate a decision forest on the training data, we can achieve multiple sets of decision trees with the combination of sensible logic rules and a higher accuracy rate due to the majority voting method [ 57 ]. Decision forest classifiers [ 54 , 68 ] are the type of ensemble methods that are used frequently. These are also used in seizure detection as they provide a high accuracy rate which depends on the majority voting method from the ensemble of decision trees. Moreover, they produce more logic rules as multiple decision trees from the training data ( D ) [ 123 ]. These logic rules are humanly interpretable, and data scientists can easily interrelate them with other seizure-related information from EEG datasets.

Siddiqui and Islam [ 125 ] used Systematic Forest (SySFor) to detect the seizure on ECoG without epoch reduction. Further, Siddiqui et al. [ 63 ] applied two decision forests—Systematic Forest (SysFor) [ 123 ] and Forest CERN [ 51 ] on nine statistical features for quick seizure detection using the concept of epoch length reduction. It is based on dividing the size of training dataset D into \(D_1, D_2\) , ... \(D_n\) and testing the accuracy at every epoch of the dataset. These sub-datasets are in descending order in terms of time duration. If the seizure can be detected in a shorter epoch length without a decline in accuracy, then we can use the same one, which results in fast seizure detection. They achieved 100% accuracy. The limitation of this work is that authors have taken the dataset of a single patient, this could be tested for more patients. Several researchers have taken the advantages of random forest classifier for detecting the seizures [ 76 , 78 , 82 , 126 ]. Because researchers/data scientists are able to see the logic rules and interpret them correspondingly. Moreover, it also provides good accuracy [ 44 , 76 , 77 , 78 , 80 , 82 ]. Donos et al. [ 44 ] applied decision forest classifier—random forest, on time and frequency domains’ feature, which was extracted from an IEEG (Intra-cranial EEG) dataset. It helped in selecting the intra-cranial channels for early seizure detection in a closed-loop circuit. The results claimed that the system can detect the seizure with 93.8% sensitivity. Wang et al. [ 94 ] developed the greedy approach of random forest, i.e., forest-grid search optimization (RF-GSO), with this method and they found 96.7% accuracy. The shortcoming of this technique is that the performance could decline if EEG signals are too noisy. Tzimourta et al. [ 93 ] applied random forest to monitor seizure activities on the two benchmark epilepsy datasets [ 64 , 65 ], the reported performance is 99.74%. Pinto-Orellana and Fábio R. Cerqueira [ 76 ] also used the random forest on the processed CHB-MIT dataset by a Spectro-temporal feature, and 70s, and the accuracy of each block is 98.30%.

Truong ND et al. [ 82 ] had carried out novel work of channel selection whilst detecting the seizure. Their key contribution is that they also focus on channels contributing mostly to automatic seizure detection. They used the random forest to solve channel selection and seizure detection, and which achieving 96.94% area under the curve (AUC). In another work, Mursalin et al. [ 80 ] proposed a method for seizure detection by selecting features with an Improved Correlation-based Feature Selection(ICFS). Basically it is a fusion of time and frequency domain. Then, a random forest classifier was applied for the seizure detection model. The obtained average classification accuracy by this approach was 98.75%.

Some other works have used an ensemble of ‘non-black-box’ classifiers such as boosting, bagging and random subspace [ 78 , 127 ]. Yan et al. [ 78 ] applied a boosting classifier achieving 94.26% of accuracy, although the results were not as impressive as the ones obtained by [ 44 ], which used a random forest classifier. Hosseini [ 128 ] used Random subspace classifier along with an SVM classifier, to classify and detect seizures. Here, the benefit of applying a subspace on big datasets is to divide them into sub-datasets based on the random subspace concept, and then the SVM classifier was applied to each sub-dataset. Ensemble accuracy (EA) was calculated by the majority voting method, which was 95%. Apart from this study, the same authors of Hosseini et al. [ 126 ] recently did another research using an ensemble of classifiers. First, they created bootstrap samples using a random subspace method, and then applied classifiers such as SVM, KNN, extended nearest neighbor (ENN), and multilayer perceptron (MLP) obtaining 97% accuracy. Hussein et al. [ 100 ], proposed a novel feature extraction method, i.e., L1-penalized robust regression (L1PRR), which uses three common symptoms during seizures—muscles artifacts, eyes movement, and white noise. Inputting these features help the random forest classifier to obtain 100% accuracy.

Observations In comparison to decision trees, decision forest classifiers are tremendously used on brain datasets for exploring different research goals. It is difficult to suggest a particular classifier whilst dealing with a high-dimensional dataset, but a random forest classifier can be a capable classifier. However, it also criticizes that not all the ‘non-black-box’ classifiers are peculiar to detect seizures and have also pointed out the objection on the drawback of using a single decision tree classifier.

5.4 Seizure detection based on black-box and non-black-box machine learning classifiers

From the literature, it is found that just a single machine learning classifier is not sufficient. Therefore, to take advantage of both ‘black-box’ and ‘non-black-box’ classifiers, some researchers utilized them in their experiments. This section provides a comprehensive review of classifiers applied together to detect the seizure.

Acharya et al. [ 111 ] used the ensemble of seven different classifiers—Fuzzy surgeon classifier (FSC), SVM, KNN, Probabilistic neural network, GMM, decision tree and Naïve Bayes for distinguishing the three states of a patient as ‘normal, ‘pre-ictal’ and ‘ictal’. The overall accuracy is 98.1%. Fergus et al. [ 83 ] also used distinct classifiers such as linear discriminant analysis (LDA), quadratic discriminant classifier (QDC), logistic classifier, uncorrelated normal density-based classifier (UDC), polynomial classifier, KNN, PARZEN, SVM, and decision tree on the processed data with seven features such as entropy, RMS, skewness, and variance. They contributed that the detected patient is suffering from a ‘Generalize seizure’ (means affecting whole brain region) across different patients without prior information about the seizure focal points. Mursalin et al. [ 101 ] proposed a method to reduce the data size, statistical sampling technique called optimum sample allocation technique, and to reduce the features they develop a feature selection algorithm. The analysis was done on the combination of five classifiers—SVM, KNN, NB, Logistic Model Trees (LMT) and Random forest.

Rand and Sriram [ 95 ] used four classifiers such as SVM, KNN, random forest, and Adaboost on a high-dimensional dataset prepared by 28 features. Their result shows that SVM outperforms on the cubic kernel. In another study, Manzouri et al. [ 98 ] used SVM and random forest on the dataset produced by 10-time and frequency features. In comparison to SVM-based detector, random forest classifier outperforms. Subasi et al. [ 96 ] achieved 100% of accuracy using four machine learning classifiers such as ANN, KNN, SVM, and random forest on two popular datasets—Freiburg and CHB-MIT to classify the three different states of seizures ‘pre-ictal’, ‘ictal’, and ‘inter-ictal’. Sharma et al. [ 102 ] proposed an automated system using iterative filtering and random forest for classifying the EEG signals. This work achieved classification accuracies of 99.5% on BONN dataset (A-E), for A versus E subsets, 96% for D versus E subsets, and 98.4% for ABCD versus E classes of EEG signals. Birjandtalab et al. [ 77 ] used two classifiers for different purposes; KNN is used to discriminate the ‘seizure’ and ‘non-seizure’ classes, whereas random forest is used to explore the significant channels. Here, the random forest also helps in the dimension reduction problem. The main benefit of selecting suitable channels is that it helps in providing relevant required information from the chosen channels, and reduces the computational cost of a classifier too. However, the authors did not mention here the important information from channel selection like finding the seizure location from the brain scalp. The main critic in [ 95 , 98 , 101 ] is that because of a large number of features, the attribute size of dataset will increases, and as a result the accuracy and computation time suffer.

5.4.1 Observations

We observe that some work used an ensemble of distinguished classifiers to take the benefits separately. For example, influential channel selection can be independently done using decision forest classifiers like a random forest. But authors used other classifiers such as SVM and KNN for classifying the seizure records with good accuracy.

6 Seizure localization

After a successful seizure detection, localization is an essential task for epileptic surgery [ 129 , 130 , 131 ]. Typically, localized seizures can be cured by surgery which arises either from the left or right region of the brain. The seizure monitoring tools such as ECoG and EEG are prominently helpful to identify the seizure location. The electrodes/channels are implanted in a non-invasive (for EEG) and an invasive manner (for ECoG). Their positioning is based on the 10/20 (10–20) International system, which helps in identifying the seizure location [ 132 ]. The concept of seizure localization means identifying the region of the brain affected by a seizure. Though some types of seizures such as ‘tonic-clonic’ are cured by anti-epileptic drugs (AED), patients with partial seizures in some cases might go for surgery [ 13 ]. To solve this problem, finding the seizure location is an essential and challenging task for neurologists and neurosurgeon [ 129 , 130 ]. The surgical target is to find a point/location/focal area from where a seizure is originating. The 10–20 positioning system gives some clues for identifying the location of a seizure. Recently, computational and machine learning methods have been applied to identify a seizure location [ 130 , 133 ].

Acar et al. [ 133 ] used trucker and non-linear multi-way Trucker kernels, and claimed that other classifiers such as SVD and principal component analysis (PCA) were unable to localize a seizure. Ghannad-Rezaie [ 134 ] applied an advanced swarm intelligence algorithm to seizure data for finding seizure location. Their study produced some appreciable results, and explored whether the patient’s temporal lobe was affected by a seizure or not. They also suggested that SVM might be able to detect the seizure location. Moreover, they also focused on the reduction of ECoG electrodes. Mansouri et al. [ 135 ] proposed an algorithm for Seizure localization, which was tested on 10 sec of EEG dataset from Karuniya University. Here, they have taken the small-size dataset, because recording usually takes several hours. If they had tested on a big dataset, it would have been much better. Fakhraei et al. [ 130 ] calculated the sensitivity of each region of the brain. The confident prediction rate (CPR) was compared with the AUC of ROC plots obtained by six classifiers from the dataset of 79 patients (31 males, 48 females) with 197 medical features. The study found that CPR was more suitable than ROC. They also explored that 43 patients had the temporal lobe epilepsy (TLE) on their left sides whilst 36 patients had it on the right sides of their brains. Likewise, Rai et al. [ 136 ] proposed a method for identifying the focal points of the seizure by applying two entropy-based features—‘renyi entropy’ and ‘negentropy’ with the neural network classifier. Siddiqui et al. [ 63 ] localize the seizure using two decision forest classifiers, and their results showed that the left hemisphere of a brain was more affected by the seizures.

Observation

It is found that compared to seizure detection, machine learning classifiers have not been extensively applied for seizure localization. But some literature exist on this problem. In these reported works, authors did not mention the percentage of the affected region of the brain by a seizure, and they were not able to identify the exact location at the lobes such as occipital, frontal, parietal left and parietal right. Although, it is not our primary objective in this review paper, whilst discussing the related published research, we found some interesting clues for seizure localization.

7 Problems identified in existing literature

One of the most significant and decisive steps is to select suitable statistical features because each channel or electrode implanted on the brain provides different statistical measures. Undoubtedly, earlier researchers made their consistent efforts to find the best features. Whilst some researchers used many features [ 34 , 79 ], the others applied a few features [ 31 , 36 , 108 , 112 , 137 ] for detecting the seizure. As a data scientist, it is very important to see the different statistical perspectives of each brain signal by analyzing the statistical properties of the features such as entropy, energy, and skewness. And we must not focus on taking irrelevant feature(s) as such since it will unnecessarily increase the dataset size. Consequently, it will be more a burden to machine learning classifiers than a benefit, and if we take few features as previous researchers did [ 71 , 73 , 79 ], this will give the low-dimensional dataset and it will not be beneficial for an effective knowledge discovery process. Therefore, we should select those potential features that can to provide logical results. Hence, it is advisable to select a group of features to avoid a burden to the machine learning classifiers and to get help in related knowledge discovery.

Each classifier has its own merits and demerits, depending on the dataset attributes and requirements [ 138 ]. In general, it is very difficult to point out which classifier was the most effective for brain datasets. To identify the capable classifier, several classifiers have been tested on EEG datasets and their performance has been evaluated, and the one which performs well is to be considered in solving seizure detection and imparting knowledge discovery. The literature reveals that previous researchers had applied different approaches, most of which were from ‘black-box’ such as ANN, KNN and SVM. The biggest shortcoming in them is that they are unable to provide the appropriate explanations for patterns and the logic rules hidden inside the models. That is why, they are not suggested for remarkable knowledge discovery process. Data scientists may not explore the internal processing of patterns [ 51 , 104 ]. However, from the literature, it is noted that the ‘non-black-box’ approach, especially, random forest, is widely used for seizure detection [ 44 , 76 , 77 ], because of its nature of generating bootstrap samples [ 124 , 139 ] whilst building a decision forest. An analysis has been done to estimate the performance of machine learning classifiers on EEG datasets and has been found that ensemble non-black-classifiers performs effectively [ 104 ]. We argue that the random forest is based on bootstrap samples and it misses some influential attributes, because it randomly selects the attribute and sometimes generates the same set of logic rules also. As a result, sometimes, it creates irrelevant information too. To overcome this issue, we also suggest some other decision forest algorithms such as SysFor [ 123 ] and Forest CERN [ 51 ] methods in seizure detection.

All these findings on seizure detection raise few interesting research questions such as selecting suitable statistical features and machine learning classifiers to take less computation time as dataset has a high volume with high dimension, and the most significant missing information from machine learning classifiers is locating the accurate point of seizure at the brain lobe(s).

7.1 Class imbalance issue in seizure detection

Class imbalance is one of the serious problems [ 140 ] in machine learning and the majority is seen in medical datasets [ 141 ], particularly in EEG signals. This is because the duration of EEG recording is long, time-consuming and seizure duration is for a few seconds, which results in being prone to errors [ 91 ]. As a result, the dataset becomes highly imbalanced. Previous researchers have focused on seizure detection. Over the last few years, researchers have been focusing on the class imbalance challenge whilst detecting the seizures, and attempting to solve it by applying different conventional approaches with some novelties. Javad Birjandtalab et al. [ 91 ] used ANN with a weighted cost function to imbalanced EEG dataset, by achieving 86% F-measure. El Saadi et al. [ 142 ] obtained 97.3% accuracy using the under-sampling method with the SVM classifier. In another work by Saadullah and Awais [ 143 ], they used a combination of SMOTE and RUSTBOST techniques for detecting seizure to imbalance seizure data with 97% accuracy. However, the research done by Yuan Qi et al. [ 86 ] was very close to the satisfactory result as they assigned the heavy weights to a minority class of the data to maintain the effective balance and solved the biasing issue. The main critique of this work is that the authors did not mentioned what weights were assigned and what was their threshold level? Here, we argue that despite of EEG data are highly imbalanced as a result of their long-hour EEG recordings, the recordings continue until the seizure is detected. The seizure(s) time spans from only seconds to minute(s). Although researchers [ 76 , 86 , 117 , 143 ] made their efforts in addressing this issue using both ‘black-box’ and ‘non-black-box’ classifiers, they did not propose any justifiable solutions, in terms of how big weights should be assigned to the minority (seizure) classes.

8 Overall observation about capable classifiers and statistical features

It is challenging to suggest that a specific classifier should be capable for seizure detection. If we discuss classifiers, three constraints are very important whilst selecting a classifier—able to handle the high-dimensional dataset, high accuracy of the model, and able to retrieve the sensible knowledge. Not all machine learning classifiers are suitable for seizure detection and knowledge discovery tasks, mainly because of their black-box nature. This means that the logic rules/patterns are not visible and understandable to data scientists. In ‘non-black-box’ classifiers amongst decision trees [ 53 ] and decision forests [ 54 ], only decision forest algorithms are more capable, because the logic rules and knowledge discovered by a single decision tree are often limited and insufficient. For example, if we build a decision tree on a training dataset—it provides a limited or single set of logic rules and stops growing the tree further as all the data points in the training set accept that rule. On the other hand, if we build a decision forest on the same training set, we get multiple decision trees with more sensible logic rules. Siddiqui et al. [ 104 ] have done the analysis on CHB-MIT dataset to know which classifier performs better. For this, they applied two black-box (SVM and KNN) and two non-black-box (decision tree and ensemble of trees i.e., bagging, random subspace, boosting); they found non-black box classifier (ensemble) outperforms compared to other classifiers of black-box. Even ensemble also performs better than a single decision tree which is a non-black box classifier. Siddiqui et al. [ 63 ] applied two decision forests—Systematic Forest (SysFor) and Forest CERN for quick seizure detection using the concept of epoch length reduction. They achieved 100% of accuracy. Similarly, Hussein et al. [ 100 ] also achieved 100% accuracy using decision forest–random forest approach.

The literature reveals that in the last few years, ‘non-black-box’ classifiers, particularly decision forest approach, were widely used on brain datasets of EEG and ECoG for different research goals [ 76 , 82 , 94 , 144 ]. The reasons for using the decision forest for seizure detection are as follows:

A decision forest overcomes some of the disadvantages of a decision tree. A decision tree discovers only a single set of logic rules from an input dataset. The logic rules that are discovered by a single decision tree may fail to correctly predict and classify the class values;

A decision forest can produce more set of logic rules/patterns compared to a single decision tree and there is a high chance of good prediction/classification compared to a single decision tree;

Able to handle high-dimensional sets;

Due to its ensemble nature a decision forest mostly produces a high accuracy compared to a single tree and other classifiers [ 54 ];

Less computational time (specifically for Random forest);

Logic rules are clear and humanly interpretable such as analysts/domain experts can easily understand and suggest best opinions. For example, affected brain lobe by seizure, identifying suitable statistical features, etc.

Furthermore, many statistical features have been used for seizure detection. However, a comparison between them is difficult because of their heterogeneous nature. Some researchers used a single feature such as energy and entropy. On the other hand, a combination of statistical features such as energy, kurtosis, line length, entropy, skewness, max, standard deviation, and min may produce promising outcomes. Most research [ 34 , 46 , 92 , 100 , 109 , 145 ] have achieved better results using these features. The novelty of [ 29 , 63 , 104 , 125 ] is the selected nine statistical features are able to assist in seizure detection with high accuracy, i.e., 100%. This also provides the clue about seizure localization with the help of sensible logical rules. Hence, the selected group of features will not be a burden to the machine learning classifier but it will assist in related knowledge discovery.

9 Research directions in seizure detection

In this research analysis, we surveyed different machine learning classifiers used for seizure detection. No doubt, the progress of the persistent attempt has been found in this topic but few interesting research questions are also raised. In this section, we identify significant challenges which can uplift the future research in this area.

Selecting suitable statistical features and machine learning classifiers to take less computation time as the dataset has a high volume with a high dimension.

Accurate seizure detection on imbalanced datasets of long duration EEG recording datasets.

Quick seizure detection on long-hour EEG recording.

Whilst selecting the machine classifier it should be kept in mind that the classifier does not miss any necessary EEG channel/electrode.

Knowledge discovery from machine learning classifiers such as seizure localization which exactly points affected brain lobe(s), channel importance, and based on participating channels in seizure a knowledge could be provided to neurologist or neurosurgeon for suggesting epilepsy category.

10 Conclusion

With the increase of epilepsy, its accurate detection becomes increasingly important. A major challenge is to detect seizures correctly from a large volume of data. Due to the complexity of EEG signals in such datasets, machine learning classifiers are suitable for accurate seizure detection. Selecting suitable classifiers and features are, however, crucial.

As such, this paper has comprehensively reviewed machine learning approaches for seizure detection. As a result, we conclude that ‘non-black-box’ classifiers—decision forest (ensemble of decision trees)—is most effective. This is because it can produce multiple sensible, explanatory logic rules with high accuracy of prediction. Further, it can help discover some relevant information such as seizure localization and exploring seizure types. On the contrary, ‘black-box’ classifiers cannot generate logic rules, although they can achieve high predictive accuracy. As for selecting suitable features, we should select those that can provide logical results. By the review of the literature, the use of the features such as entropy, line length, energy, skewness, kurtosis, and standard deviation can achieve 100% accuracy in the classifiers. We suggest not to use the irrelevant features as the dimension of the data increases. This is because the computation cost of a classifier will grow high, and it may also produce insensible patterns. If we use just one or two features such as line length and energy, the low-dimensional dataset will be generated. However, this dataset will not be fruitful for the knowledge discovery process.

This review paper has provided new perspectives to data scientists who are working on epileptic seizure detection using EEG signals. In summary, this paper focuses on the review of selecting machine learning classifiers and suitable features.

Availability of data and materials

Not applicable.

Organization WH (2006) Neurological disorders: public health challenges. World Health Organization, New York

Google Scholar  

Chaudhary UJ, Duncan JS, Lemieux L (2011) A dialogue with historical concepts of epilepsy from the babylonians to hughlings jackson: persistent beliefs. Epilepsy Behav 21(2):109–114

Article   Google Scholar  

Reynolds EH (2009) Milestones in epilepsy*. Epilepsia 50(3):338–342. https://doi.org/10.1111/j.1528-1167.2009.02050.x

WHO (2005) Atlas: Epilepsy care in the world. World Health Organization, Geneva

Alarcón G, Valentín A (2012) Introduction to Epilepsy. Cambridge University Press, Cambridge

Book   Google Scholar  

Qu H, Gotman J (1993) Improvement in seizure detection performance by automatic adaptation to the eeg of each patient. Electroencephalogr Clin Neurophysiol 86(2):79–87

Sazgar M, Young MG (2019) Seizures and Epilepsy. In: Absolute Epilepsy and EEG rotation review, pp. 9–46. Springer, New York

Chapter   Google Scholar  

Schachter S, Shafer P, Sirven J (2013) What causes epilepsy and seizures. Epilepsy Foundation

Delanty N, Vaughan CJ, French JA (1998) Medical causes of seizures. Lancet 352(9125):383–390

WHO: Media Center Epilepsy, (Fact sheet N999). http://www.who.int/mediacentre/factsheets/fs999/en/ (2015) Accessed 15 July 2019

Shafer PO, Sirven JI (2014) Epilepsy statistics. Epilepsy Foundation

Hannah JA, Brodie MJ (1998) Epilepsy and learning disabilities—a challenge for the next millennium? Seizure 7(1):3–13

Thurman DJ, Beghi E, Begley CE, Berg AT, Buchhalter JR, Ding D, Hesdorffer DC, Hauser WA, Kazis L, Kobau R et al (2011) Standards for epidemiologic studies and surveillance of epilepsy. Epilepsia 52(s7):2–26

Fisher RS (2017) The new classification of seizures by the international league against epilepsy 2017. Curr Neurol Neurosci Rep 17(6):48

Mahmud M, Kaiser MS, Hussain A, Vassanelli S (2018) Applications of deep learning and reinforcement learning to biological data. IEEE Trans Neural Networks Learn Syst 29(6):2063–2079

Article   MathSciNet   Google Scholar  

Mahmud M, Kaiser MS, Hussain A (2020) Deep learning in mining biological data. arXiv:2003.00108

Fayyad UM, Piatetsky-Shapiro G, Smyth P, Uthurusamy R (eds.) (1996) Advances in Knowledge Discovery and Data Mining. American Association for Artificial Intelligence, Menlo Park, CA, USA

Yin Y, Kaku I, Tang J, Zhu J (2011) Data mining: concepts, methods and applications in management and engineering design. Springer, New York

Islam MZ, D’Alessandro S, Furner M, Johnson L, Gray D, Carter L (2016) Brand switching pattern discovery by data mining techniques for the telecommunication industry in australia. Aust J Inform Syst. https://doi.org/10.3127/ajis.v20i0.1420

Aljumah AA, Ahamad MG, Siddiqui MK (2013) Application of data mining: diabetes health care in young and old patients. J King Saud Univ Comput Inform Sci 25(2):127–136

Aljumah A, Siddiqui M (2016) Data mining perspective: prognosis of life style on hypertension and diabetes. Int Arab J Inf Technol 13(1)

Siddiqui MK, Morales-Menendez R, Gupta PK, Iqbal HM, Hussain F, Khatoon K, Ahmad S (2020) Correlation between temperature and COVID-19 (suspected, confirmed and death) cases based on machine learning analysis. J Pure Appl Microbiol 14(Spl Edn.)

Aljumah AA, Siddiqui MK (2014) Hypertension interventions using classification based data mining. Res J Appl Sci Eng Technol 7(17):3593–3602

Almazyad AS, Ahamad MG, Siddiqui MK, Almazyad AS (2010) Effective hypertensive treatment using data mining in Saudi Arabia. J Clin Monit Comput 24(6):391–401

Singh GA, Gupta PK (2019) Performance analysis of various machine learning-based approaches for detection and classification of lung cancer in humans. Neural Comput Appl 31(10):6863–6877

Fu T-c (2011) A review on time series data mining. Eng Appl Artif Intell 24(1):164–181

Tzallas AT, Tsipouras MG, Tsalikakis DG, Karvounis EC, Astrakas L, Konitsiotis S, Tzaphlidou M (2012) Automated epileptic seizure detection methods: a review study. In: Epilepsy-histological, electroencephalographic and psychological aspects. InTech

Abbasi B, Goldenholz DM (2019) Machine learning applications in epilepsy. Epilepsia

Siddiqui MK. Brain data mining for epileptic seizure-detection. Doctoral Dissertation, Charles Sturt University, Australia

Paul Y (2018) Various epileptic seizure detection techniques using biomedical signals: a review. Brain Inform 5(2):6

Boonyakitanont P, Lek-uthai A, Chomtho K, Songsiri J (2020) A review of feature extraction and performance evaluation in epileptic seizure detection using eeg. Biomed Signal Process Control 57:101702

Sharmila A, Geethanjali P (2019) A review on the pattern detection methods for epilepsy seizure detection from EEG signals. Biomed Eng /Biomedizinische Technik. 64(5):507–17

van Westrhenen A, De Cooman T, Lazeron RH, Van Huffel S, Thijs RD (2019) Ictal autonomic changes as a tool for seizure detection: a systematic review. Clin Auton Res 29(2):161–181

Logesparan AJC Lojini, Rodriguez-Villegas E (2012) Optimal features for online seizure detection. Med Biol Eng Comput 50(7):659–669

Ramgopal S, Thome-Souza S, Jackson M, Kadish NE, Fernández IS, Klehm J, Bosl W, Reinsberger C, Schachter S, Loddenkemper T (2014) Seizure detection, seizure prediction, and closed-loop warning systems in epilepsy. Epilepsy Behav 37:291–307

Quintero-Rincón A, D’Giano C, Batatia H (2019) Seizure onset detection in eeg signals based on entropy from generalized gaussian pdf modeling and ensemble bagging classifier. In: Digital Health Approach for Predictive, Preventive, Personalised and Participatory Medicine, pp. 1–10. Springer, New York

Greene B, Faul S, Marnane W, Lightbody G, Korotchikova I, Boylan G (2008) A comparison of quantitative eeg features for neonatal seizure detection. Clin Neurophysiol 119(6):1248–1261

Gotman J, Ives J, Gloor P (1979) Automatic recognition of inter-ictal epileptic activity in prolonged eeg recordings. Electroencephalogr Clin Neurophysiol 46(5):510–520

Gotman J (1982) Automatic recognition of epileptic seizures in the eeg. Electroencephalogr Clin Neurophysiol 54(5):530–540

Koffler D, Gotman J (1985) Automatic detection of spike-and-wave bursts in ambulatory eeg recordings. Electroencephalogr Clin Neurophysiol 61(2):165–180

Shoeb A, Guttag J (2010) Application of machine learning to epileptic seizure detection. In: 2010 the 27th International Conference on Machinelearning, Haifa, Israel

Dorai A, Ponnambalam K (2010) Automated epileptic seizure onset detection. In: Autonomous and Intelligent Systems (AIS), 2010 International Conference On, pp. 1–4 . IEEE

Lahmiri S (2018) An accurate system to distinguish between normal and abnormal electroencephalogram records with epileptic seizure free intervals. Biomed Sign Process Control 40:312–317

Donos C, Dümpelmann M, Schulze-Bonhage A (2015) Early seizure detection algorithm based on intracranial EEG and random forest classification. Int J Neur Syst 25(05):1550023

Ocak H (2009) Automatic detection of epileptic seizures in eeg using discrete wavelet transform and approximate entropy. Exp Syst Appl 36(2):2027–2036

Thomas E, Temko A, Lightbody G, Marnane W, Boylan G (2011) Advances in automated neonatal seizure detection. New Adv Intell Sign Process 1:93–113

Al Ghayab HR, Li Y, Abdulla S, Diykh M, Wan X (2016) Classification of epileptic EEG signals based on simple random sampling and sequential feature selection. Brain Inform 3(2):85–91

Herwig U, Satrapi P, Schönfeldt-Lecuona C (2003) Using the international 10–20 eeg system for positioning of transcranial magnetic stimulation. Brain Topogr 16(2):95–99

Misulis KE (2013) Atlas of EEG, seizure semiology, and management. Oxford University Press, Oxford

Sammut C, Webb GI (2017) Encyclopedia of machine learning and data mining. Springer, New York

Book   MATH   Google Scholar  

Adnan MN, Islam MZ (2016) Forest CERN: A new decision forest building technique. In: Pacific-Asia Conference on Knowledge Discovery and Data Mining, pp. 304–315. Springer, New York

Cortes C, Vapnik V (1995) Support-vector networks. Mach Learn 20(3):273–297

MATH   Google Scholar  

Quinlan JR (1993) C4.5: Programs for machine learning. Morgan Kaufmann Publishers Inc., San Francisco

Breiman L (2001) Random forests. Mach Learn 45(1):5–32. https://doi.org/10.1023/A:1010933404324

Article   MATH   Google Scholar  

Arlot S, Celisse A et al (2010) A survey of cross-validation procedures for model selection. Statistics surveys 4:40–79

Article   MathSciNet   MATH   Google Scholar  

Kurgan LA, Cios KJ (2004) CAIM discretization algorithm. IEEE Trans Knowl Data Eng 16(2):145–153

Li J, Liu H (2003) Ensembles of cascading trees. In: Data Mining, 2003. ICDM 2003. Third IEEE International Conference On, pp. 585–588. IEEE

Powers DM (2011) Evaluation: from precision, recall and f-measure to roc, informedness, markedness and correlation

Goldberger A, Amaral L, Glass L, Hausdorff J, Ivanov PC, Mark R, Mietus J, Moody G, Peng C, Stanley H (2000) PhysioBank, PhysioToolkit, and PhysioNet: components of a new research resource for complex physiologic signals. Circulation 101(23):e215–e220

CHB-MIT Scalp EEG Database. https://physionet.org/pn6/chbmit/ . Accessed 20 Jun 2015

Shoeb A, Edwards H, Connolly J, Bourgeois B, Treves ST, Guttag J (2004) Patient-specific seizure onset detection. Epilepsy Behav 5(4):483–498

Kramer MA, Kolaczyk ED, Kirsch HE (2008) Emergent network topology at seizure onset in humans. Epilepsy Res 79(2):173–186

Siddiqui MK, Islam MZ, Kabir MA (2018) A novel quick seizure detection and localization through brain data mining on ecog dataset. Neural Computing and Applications 1–14

Freiburg seizure prediction project. Freiburg, Germany. http://epilepsy.uni-freiburg.de/freiburg-seizure-prediction-project/eeg-database (2003)

Andrzejak RG, Lehnertz K, Mormann F, Rieke C, David P, Elger CE (2001) Indications of nonlinear deterministic and finite-dimensional structures in time series of brain electrical activity: Dependence on recording region and brain state. Phys Rev E 64(6):061907

Andrzejak RG, Schindler K, Rummel C (2012) Nonrandomness, nonlinear dependence, and nonstationarity of electroencephalographic recordings from epilepsy patients. Phys Rev E 86(4):046206

Lee H, Kim S (2016) Black-box classifier interpretation using decision tree and fuzzy logic-based classifier implementation. Int J Fuzzy Logic Intell Syst 16(1):27–35

Adnan MN, Islam MZ (2017) Forex++: A new framework for knowledge discovery from decision forests. Aust J Inform Syst. https://doi.org/10.3127/ajis.v21i0.1539

Guo L, Rivero D, Dorado J, Rabunal JR, Pazos A (2010) Automatic epileptic seizure detection in eegs based on line length feature and artificial neural networks. J Neurosci Methods 191(1):101–109

Koolen N, Jansen K, Vervisch J, Matic V, De Vos M, Naulaers G, Van Huffel S (2014) Line length as a robust method to detect high-activity events: automated burst detection in premature eeg recordings. Clin Neurophysiol 125(10):1985–1994

Logesparan L, Rodriguez-Villegas E, Casson AJ (2015) The impact of signal normalization on seizure detection using line length features. Med Biol Eng Comput 53(10):929–942

Song Y, Liò P (2010) A new approach for epileptic seizure detection: sample entropy based feature extraction and extreme learning machine. J Biomed Sci Eng 3(06):556

Zhang Y, Zhang Y, Wang J, Zheng X (2014) Comparison of classification methods on EEG signals based on wavelet packet decomposition. Neural Comput Appl 26(5):1217–1225. https://doi.org/10.1007/s00521-014-1786-7

Ahmad MA, Khan NA, Majeed W (2014) Computer assisted analysis system of electroencephalogram for diagnosing epilepsy. In: Pattern Recognition (ICPR), 2014 22nd International Conference On, pp. 3386–3391 . IEEE

Gill AF, Fatima SA, Akram MU, Khawaja SG, Awan SE (2015) Analysis of eeg signals for detection of epileptic seizure using hybrid feature set. In: Theory and Applications of Applied Electromagnetics, pp. 49–57. Springer

Orellana MP, Cerqueira F (2016) Personalized epilepsy seizure detection using random forest classification over one-dimension transformed EEG data. bioRxiv, 070300

Birjandtalab J, Pouyan MB, Cogan D, Nourani M, Harvey J (2017) Automated seizure detection using limited-channel eeg and non-linear dimension reduction. Comput Biol Med 82:49–58

Yan A, Zhou W, Yuan Q, Yuan S, Wu Q, Zhao X, Wang J (2015) Automatic seizure detection using stockwell transform and boosting algorithm for long-term eeg. Epilepsy Behav 45:8–14

Amin HU, Malik AS, Ahmad RF, Badruddin N, Kamel N, Hussain M, Chooi W-T (2015) Feature extraction and classification for eeg signals using wavelet transform and machine learning techniques. Austr Phys Eng Sci Med 38(1):139–149

Mursalin M, Zhang Y, Chen Y, Chawla NV (2017) Automated epileptic seizure detection using improved correlation-based feature selection with random forest classifier. Neurocomputing 241:204–214

Zabihi M, Kiranyaz S, Ince T, Gabbouj M (2013) Patient-specific epileptic seizure detection in long-term eeg recording in paediatric patients with intractable seizures

Truong ND, Kuhlmann L, Bonyadi MR, Yang J, Faulks A, Kavehei O (2017) Supervised learning in automatic channel selection for epileptic seizure detection. Exp Syst Appl

Fergus P, Hussain A, Hignett D, Al-Jumeily D, Abdel-Aziz K, Hamdan H (2016) A machine learning system for automated whole-brain seizure detection. Appl Comput Inform 12(1):70–89

Chen D, Wan S, Xiang J, Bao FS (2017) A high-performance seizure detection algorithm based on discrete wavelet transform (dwt) and eeg. PLoS ONE 12(3):0173138

Satapathy SK, Jagadev AK, Dehuri S (2017) Weighted majority voting based ensemble of classifiers using different machine learning techniques for classification of eeg signal to detect epileptic seizure. Informatica 41(1):99

MathSciNet   Google Scholar  

Yuan Q, Zhou W, Zhang L, Zhang F, Xu F, Leng Y, Wei D, Chen M (2017) Epileptic seizure detection based on imbalanced classification and wavelet packet transform. Seizure

Subasi A, Kevric J, Canbaz MA (2019) Epileptic seizure detection using hybrid machine learning methods. Neural Comput Appl 31(1):317–325

Al Ghayab HR, Li Y, Siuly S, Abdulla S (2019) Epileptic seizures detection in eegs blending frequency domain with information gain technique. Soft Comput 23(1):227–239

Selvakumari RS, Mahalakshmi M, Prashalee P (2019) Patient-specific seizure detection method using hybrid classifier with optimized electrodes. J Med Syst 43(5):121

Chen S, Zhang X, Chen L, Yang Z (2019) Automatic diagnosis of epileptic seizure in electroencephalography signals using nonlinear dynamics features. IEEE Access 7:61046–61056

Birjandtalab J, Jarmale VN, Nourani M, Harvey J (2018) Imbalance learning using neural networks for seizure detection. In: 2018 IEEE Biomedical Circuits and Systems Conference (BioCAS), pp. 1–4 . IEEE

Lahmiri S, Shmuel A (2018) Accurate classification of seizure and seizure-free intervals of intracranial eeg signals from epileptic patients. IEEE Trans Instrum Meas 68(3):791–796

Tzimourta KD, Tzallas AT, Giannakeas N, Astrakas LG, Tsalikakis DG, Angelidis P, Tsipouras MG (2019) A robust methodology for classification of epileptic seizures in eeg signals. Health Technol 9(2):135–142

Wang X, Gong G, Li N, Qiu S (2019) Detection analysis of epileptic EEG using a novel random forest model combined with grid search optimization. Front Human Neurosci 13:12

Raghu S, Sriraam N (2018) Classification of focal and non-focal eeg signals using neighborhood component analysis and machine learning algorithms. Exp Syst Appl 113:18–32

Alickovic E, Kevric J, Subasi A (2018) Performance evaluation of empirical mode decomposition, discrete wavelet transform, and wavelet packed decomposition for automated epileptic seizure detection and prediction. Biomed Sign Process Contr 39:94–102

Fasil O, Rajesh R (2019) Time-domain exponential energy for epileptic eeg signal classification. Neurosci Lett 694:1–8

Manzouri F, Heller S, Dümpelmann M, Woias P, Schulze-Bonhage A (2018) A comparison of machine learning classifiers for energy-efficient implementation of seizure detection. Front Syst Neurosci 12: 43

Parvez MZ, Paul M (2014) Epileptic seizure detection by analyzing eeg signals using different transformation techniques. Neurocomputing 145:190–200

Hussein R, Elgendi M, Wang ZJ, Ward RK (2018) Robust detection of epileptic seizures based on l1-penalized robust regression of EEG signals. Exp Syst Appl 104:153–167

Mursalin M, Islam SS, Noman MK (2019) Epileptic seizure classification using statistical sampling and a novel feature selection algorithm. arXiv preprint arXiv:1902.09962

Sharma RR, Varshney P, Pachori RB, Vishvakarma SK (2018) Automated system for epileptic EEG detection using iterative filtering. IEEE Sens Lett 2(4):1–4

Tzallas AT, Tsipouras MG (2007) Fotiadis DI (2007) Automatic seizure detection based on time-frequency analysis and artificial neural networks. Comput Intell Neurosci 2007:80510

Siddiqui MK, Islam MZ, Kabir MA (2017) Analyzing performance of classification techniques in detecting epileptic seizure. In: International Conference on Advanced Data Mining and Applications, pp. 386–398. Springer, New York

Sanei S, Chambers JA (2013) EEG signal processing. Wiley, New York

Chaovalitwongse WA, Prokopyev OA, Pardalos PM (2006) Electroencephalogram (eeg) time series classification: applications in epilepsy. Ann Ope Res 148(1):227–250

Moselhy HF (2011) Psychosocial and cultural aspects of epilepsy. In: Novel Aspects on Epilepsy. InTech

Esteller R, Echauz J, Tcheng T, Litt B, Pless B (2001) Line length: an efficient feature for seizure onset detection. In: Engineering in Medicine and Biology Society, 2001. Proceedings of the 23rd Annual International Conference of the IEEE, vol. 2, pp. 1707–1710 . IEEE

Guerrero-Mosquera C, Trigueros AM, Franco JI, Navia-Vázquez Á (2010) New feature extraction approach for epileptic eeg signal detection using time-frequency distributions. Med Biol Eng Comput 48(4):321–330

Olsen DE, Lesser RP, Harris JC, Webber WRS, Cristion JA (1994) Automatic detection of seizures using electroencephalographic signals. Google Patents. US Patent 5,311, 876

Acharya UR, Molinari F, Sree SV, Chattopadhyay S, Ng K-H, Suri JS (2012) Automated diagnosis of epileptic EEG using entropies. Biomed Sign Process Contr 7(4):401–408

Kannathal N, Choo ML, Acharya UR, Sadasivan P (2005) Entropies for detection of epilepsy in eeg. Comput Meth Progr Biomed 80(3):187–194

Kumar Y, Dewal M (2011) Complexity measures for normal and epileptic eeg signals using apen, sampen and sen. IJCCT 2(7):6–12

Shimizu M, Iiya M, Fujii H, Kimura S, Suzuki M, Nishizaki M (2019) Left ventricular end-systolic contractile entropy can predict cardiac prognosis in patients with complete left bundle branch block. Journal of Nuclear Cardiology 1–10

Pal PR, Panda R (2010) Classification of eeg signals for epileptic seizure evaluation. In: Students’ Technology Symposium (TechSym), 2010 IEEE, pp. 72–76. IEEE

Cepukenas J, Lin C, Sleeman D (2015) Applying rule extraction and rule refinement techniques to (blackbox) classifiers. In: Proceedings of the 8th international conference on knowledge capture, p. 27. ACM

Birjandtalab J, Pouyan MB, Nourani M (2016) Unsupervised eeg analysis for automated epileptic seizure detection. In: Proceedings of the first international workshop on pattern recognition, international society for optics and photonics, pp. 100110–100110

Abualsaud K, Mahmuddin M, Saleh M, Mohamed A (2015) Ensemble classifier for epileptic seizure detection for imperfect eeg data. Sci World J 2015:945689

Chen C, Liu J, Syu J (2012) Application of chaos theory and data mining to seizure detection of epilepsy. Proc Conf. IPCSIT/Hong Kong 25:23–28

Polat K, Güneş S (2007) Classification of epileptiform eeg using a hybrid system based on decision tree classifier and fast fourier transform. Appl Math Comput 187(2):1017–1026

MathSciNet   MATH   Google Scholar  

Quinlan R (2004) Data mining tools see5 and c5. 0

Cios KJ, Pedrycz W, Swiniarski RW, Kurgan LA (2007) Data mining: a knowledge discovery approach. Springer, New York

Islam MZ, Giggins H (2011) Knowledge discovery through sysfor: a systematically developed forest of multiple decision trees. In: Proceedings of the Ninth Australasian Data Mining Conference-Volume 121, pp. 195–204. Australian Computer Society, Inc

Ho TK (1998) The random subspace method for constructing decision forests. IEEE Trans Patt Analy Mach Intell 20(8):832–844

Siddiqui MK, Islam MZ (2016) Data mining approach in seizure detection. In: 2016 IEEE Region 10 Conference (TENCON), Singapore, pp. 3579–3583. Institute of Electrical and Electronics Engineers (IEEE), https://doi.org/10.1109/tencon.2016.7848724

Hosseini M-P, Pompili D, Elisevich K, Soltanian-Zadeh H (2018) Random ensemble learning for eeg classification. Artif Intell Med 84:146–158

Hassan AR, Siuly S, Zhang Y (2016) Epileptic seizure detection in eeg signals using tunable-q factor wavelet transform and bootstrap aggregating. Comput Meth Prog Biomed 137:247–259

Hosseini M-P, Hajisami A, Pompili D (2016) Real-time epileptic seizure detection from eeg signals via random subspace ensemble learning. In: Autonomic Computing (ICAC), 2016 IEEE International Conference On, pp. 209–218 . IEEE

Jurcak V, Tsuzuki D, Dan I (2007) 10/20, 10/10, and 10/5 systems revisited: Their validity as relative head-surface-based positioning systems. NeuroImage 34(4):1600–1611. https://doi.org/10.1016/j.neuroimage.2006.09.024

Fakhraei S, Soltanian-Zadeh H, Fotouhi F, Elisevich K (2011) Confidence in medical decision making: Application in temporal lobe epilepsy data mining. In: Proceedings of the 2011 Workshop on Data Mining for Medicine and Healthcare. DMMH ’11, pp. 60–63. ACM, New York, NY, USA . https://doi.org/10.1145/2023582.2023593

Keraudren K, Kainz B, Oktay O, Kyriakopoulou V, Rutherford M, Hajnal JV, Rueckert D (2015) Automated localization of fetal organs in mri using random forests with steerable features. In: International conference on medical image computing and computer-assisted Intervention, pp. 620–627. Springer, New York

Trans-Cranial-Technologies: 10/20 System Positioning. https://www.trans-cranial.com/local/manuals/10_20_pos_man_v1_0_pdf.pdf . Accessed 7 Oct 2015

Acar E, Bingöl CA, Bingöl H, Yener B (2006) Computational analysis of epileptic focus localization. In: Proceedings of the 24th IASTED International Conference on Biomedical Engineering. BioMed’06, pp. 317–322. ACTA Press, Anaheim, CA, USA. http://dl.acm.org/citation.cfm?id=1166506.1166562

Ghannad-Rezaie M, Soltanain-Zadeh H, Siadat MR, Elisevich KV (2006) Medical data mining using particle swarm optimization for temporal lobe epilepsy. In: 2006 IEEE International conference on evolutionary computation, pp. 761–768. https://doi.org/10.1109/CEC.2006.1688388

Mansouri A, Singh SP, Sayood K (2019) Online eeg seizure detection and localization. Algorithms 12(9):176

Rai K, Bajaj V, Kumar A (2015) Features extraction for classification of focal and non-focal eeg signals. Inform Sci Appl Lect Notes Electr Eng 339:599–605

Hadoush H, Alafeef M, Abdulhay E (2019) Brain complexity in children with mild and severe autism spectrum disorders: Analysis of multiscale entropy in eeg. Brain topography 1–8

Fernández-Delgado M, Cernadas E, Barro S, Amorim D (2014) Do we need hundreds of classifiers to solve real world classification problems? J Mach Learn Res 15(1):3133–3181

Breiman L (1996) Bagging predictors. Mach Learn 24(2):123–140

Galar M, Fernandez A, Barrenechea E, Bustince H, Herrera F (2012) A review on ensembles for the class imbalance problem: bagging-, boosting-, and hybrid-based approaches. IEEE Trans Syst Man Cybern C 42(4):463–484

Yang F, Wang H-Z, Mi H, Cai W-W et al (2009) Using random forest for reliable classification and cost-sensitive learning for medical diagnosis. BMC Bioinform 10(1):22

El Saadi H, Al-Sadek AF, Fakhr MW (2012) Informed under-sampling for enhancing patient specific epileptic seizure detection. Int J Comput Appl 57:16

Amin S, Kamboh AM (2016) A robust approach towards epileptic seizure detection. In: Machine Learning for Signal Processing (MLSP), 2016 IEEE 26th International Workshop On, pp. 1–6 . IEEE

Yang Z, Choupan J, Reutens D, Hocking J (2015) Lateralization of temporal lobe epilepsy based on resting-state functional magnetic resonance imaging and machine learning. Front Neurol 6:184

Tito M, Cabrerizo M, Ayala M, Jayakar P, Adjouadi M (2009) Seizure detection: an assessment of time-and frequency-based features in a unified two-dimensional decisional space using nonlinear decision functions. J Clin Neurophysiol 26(6):381–391

Download references

Acknowledgements

We acknowledge the contribution of Dr. Khudeja Khatoon, MD, Faculty member of the Hayat Unani Medical College & Research Centre, India for carefully looking the medical terminologies in the paper. Also, thankful to Mr. Mohammad Arshad, English language expert, Shoumou Investment and Trading Company, KSA for proofreading the paper.

Author information

Mohammad Khubeb Siddiqui and Xiaodi Huang contributed equally to this work

Authors and Affiliations

School of Engineering and Sciences, Tecnologico de Monterrey, Av. E. Garza Sada 2501, Monterrey, Nuevo Leon, Mexico

Mohammad Khubeb Siddiqui & Ruben Morales-Menendez

School of Computing and Mathematics, Charles Sturt University, 2640, Albury, NSW, Australia

Xiaodi Huang

College of Applied Studies and Community Service, King Saud University, Riyadh, Kingdom of Saudi Arabia

Nasir Hussain

You can also search for this author in PubMed   Google Scholar

Contributions

MKS: acquisition of related works, machine learning study, comparative study, arguments, writeup. RMM reviewed the overall paper and provided comments. XH: review the article on Machine Learning perspective, main contributions and direction of research. NH: discussion and table analysis. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Ruben Morales-Menendez .

Ethics declarations

Competing interests.

The authors declare that they have no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Siddiqui, M.K., Morales-Menendez, R., Huang, X. et al. A review of epileptic seizure detection using machine learning classifiers. Brain Inf. 7 , 5 (2020). https://doi.org/10.1186/s40708-020-00105-1

Download citation

Received : 24 December 2019

Accepted : 09 May 2020

Published : 25 May 2020

DOI : https://doi.org/10.1186/s40708-020-00105-1

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Applications of machine learning on epilepsy
• Statistical features
• Seizure detection
• Seizure localization
• Black-box and non-black-box classifiers
• EEG signals

REVIEW article

Epileptic seizure detection and experimental treatment: a review.

• 1 Department of Computer Science, University of Colorado, Boulder, CO, United States
• 2 Department of Computer Science and Engineering, University of Texas at Arlington, Arlington, TX, United States
• 3 Department of Computer Science, University of Oxford, Oxford, United Kingdom

One-fourths of the patients have medication-resistant seizures and require seizure detection and treatment continuously to cope with sudden seizures. Seizures can be detected by monitoring the brain and muscle activities, heart rate, oxygen level, artificial sounds, or visual signatures through EEG, EMG, ECG, motion, or audio/video recording on the human head and body. In this article, we first discuss recent advances in seizure sensing, signal processing, time- or frequency-domain analysis, and classification algorithms to detect and classify seizure stages. Then, we show a strong potential of applying recent advancements in non-invasive brain stimulation technology to treat seizures. In particular, we explain the fundamentals of brain stimulation approaches, including (1) transcranial magnetic stimulation (TMS), (2) transcranial direct current stimulation (tDCS), (3) transcranial focused ultrasound stimulation (tFUS), and how to use them to treat seizures. Through this review, we intend to provide a broad view of both recent seizure diagnoses and treatments. Such knowledge would help fresh and experienced researchers to capture the advancements in sensing, detection, classification, and treatment seizures. Last but not least, we provide potential research directions that would attract seizure researchers/engineers in the field.

1. Introduction

Epileptic seizure is a transient occurrence of signs or symptoms due to abnormal excessive or synchronous neuronal activity in the brain ( 1 ). Currently, about 2.3 million adults and more than 450,000 children and adolescents in the United States live with epilepsy. About 150,000 people are diagnosed with epileptic seizures each year ( 2 ). Epileptic seizures all start in the brain with sudden abnormal electrical discharges 1 . Among patients with epileptic seizures, two-thirds can control seizures through anti-epileptic medication, and another 8-10% could benefit from surgery. The remaining 25% have medication-resistant epileptic seizures and experience sudden seizure symptoms ( 3 ). Therefore, it is essential to notify the patient's medication-resistant epileptic seizure to the caretaker and analyze the pattern of related signals before, during, and after the seizure onset.

This article contributes to organizing seizure detection, classification, and treatment. We also provide potential research directions that would attract seizure researchers/engineers in the field. The existing seizure surveys reviewed seizure detection ( 4 ), classification ( 5 – 8 ), or treatment ( 9 , 10 ). This paper discusses state-of-the-art techniques for (1) capturing the physiology signals of seizures, (2) detecting and classifying types of seizures, (3) seizures therapy, and (4) the challenges and potential seizure-related research directions.

First, accurately and reliably capturing physiology signals related to seizure is a critical step for designing robust seizure detection systems. Monitoring brain activity signal (e.g., Electroencephalogram, EEG) is the most common method to detect seizures. The EEG recording of patients with epileptic seizures has two categories of abnormal activity: interictal, abnormal signals recorded between epileptic seizures, and ictal, the activity recorded during an epileptic seizure ( 6 ). We focused on epileptic seizure detection and considered interictal and ictal EEG signals except postictal state to detect abnormal EEG signals. The EEG signature of an inter-ictal activity is occasional transient waveforms, while that of an ictal activity is composed of a continuous discharge of polymorphic waveforms of variable amplitude and frequency ( 11 ). There are two kinds of traditional EEG recording techniques: Invasive EEG and scalp EEG. The invasive EEG recording is necessary to do surgery to implant the electrodes in the brain. In the case of the scalp EEG, the user is required to attach multiple electrodes that are connecting to a monitoring device through many wires. Therefore, Patients have to suffer the inconvenience of inserting something into the body or attaching multiple electrodes. Also, for the scalp EEG, a trained physician does such a complicated setup, and the studies are often conducted in hospitals. Besides the traditional EEG-based approach, epileptic seizures can also be detected through eye (lid) movement, heart rate, blood pressure, arterial oxygenation ( SpO 2 ), respiration, sweating, and so on ( 4 ). These activities can be captured from physiology signals, including Electrooculography (EOG), electrocardiography (ECG), electromyography (EMG), electrodermal activity (EDA), motion, audio/video recording, and multimodality sensing approaches ( 4 , 7 ).

We also discuss in detail the key components of these state of the art systems to provide a detailed picture of recent efforts on extracting these physiological signals for seizure detection. These systems often include some essential components as following: (1) signal acquisition, (2) signal processing, (3) feature extraction. The signal acquisition component is designed to capture physiological signals that are directly or indirectly related to seizures ( 4 , 12 ). These signals often contain a lot of noises, which will be processed further using novel, yet complex algorithms to extract the signal of interests ( 13 , 14 ). Next, many recent efforts have focused on building a stable setup features representing the presence of seizures to improve the detection accuracy ( 15 – 18 ). Hybrid time-frequency analysis features are often used to overcome the impact of human motion artifacts as well as to improve the system sensitivities ( 19 – 21 ). Specifically, wavelet transform analysis (WT) approaches are employed ( 22 ) to provide detailed resolutions of the seizure-related signatures on both time and frequency domains ( 23 ).

Second, after capturing the physiology signals, it is important to accurately detect and classify the type of detected seizures ( 5 , 6 , 24 ). Existing seizure classification methods primarily include classical machine learning approaches [e.g., support vector machine (SVM)] and novel deep-learning solutions [e.g., artificial neural network (ANN) ( 7 )]. SVM divides data belonging to two groups into a hyperplane ( 25 , 26 ). The original SVM is a binary classification, whereas the class for seizure is divided into at least three (focal seizure, generalized seizure, and healthy). State of the art SVM-approaches only can classify two classes of seizures (seizure vs. non-seizure) with high accuracy ( 27 , 28 ). It is not sufficient for seizure classification. Multiclass SVM methods have been used by splitting one multiclass problem into several binary classification problems ( 29 , 30 ). Although many related works have used multiclass SVM to classify various seizure types, it is impractical due to the low classification accuracy and many false alarms ( 29 , 31 , 32 ). Many recent efforts have focused on developing more complex learning algorithms. Especially, deep-learning solutions to detect a variety of seizures attract much attention from researchers ( 33 ). The classification performance depends on how the system structures hidden layers, such as multilayer perceptron neural network (MLPNN), adaptive neuro-fuzzy inference system (ANFIS), radial basis function neural network (RBFNN), convolutional neural network (CNN), and recurrent neural network (RNN) ( 34 ). ANN is the preferred method over SVM because it is not affected by the number of classes.

Third, after detecting and classifying different types of seizures, treatment methods need to be developed to reduce or remove the impact of seizures on patients' normal life. Even though it is difficult to find existing works in this direction, we believe that these can be done by exploring the uses of state of the art brain stimulation technique. We also discuss how the recent development in brain stimulation and interventions would help to treat seizures, such as decreasing cortical excitability with low-frequency magnetic stimulation ( 35 ) or counterbalancing the neuronal hyper-excitation through electric neural modulation ( 36 ). In particular, brain stimulation has been noted as an alternative to drug therapy to decrease the frequency of seizure or reduce the symptom. It is mostly divided into invasive and non-invasive. Although the invasive brain stimulation stimulates the problematic seizure part of the brain directly and provides a fast and accurate effect, it is necessary to do surgery to implant the stimulator inside the brain. It is very costly and may damage the brain during the operation. Thus, many patients are reluctant to this type of therapy. For non-invasive brain stimulation, there are two principal methods: transcranial magnetic stimulation (TMS) and transcranial direct current stimulation (tDCS) ( 10 ). TMS uses the principle of electromagnetic induction to focus induced current in the brain ( 37 ). The magnetic fields generated by TMS penetrate human tissue painlessly and induces electric currents that can depolarize neurons or their axons in the brain ( 38 ). tDCS is one of transcranial electrical stimulation (tES) and applies low-amplitude direct currents via scalp electrodes and penetrate the skull to enter the brain ( 37 ). Unlike other tES methods, tDCS delivers a sustained current ( 39 ) and can make the therapeutic effect through the sustained current. However, TMS and tDCS provide low spatial resolutions, which lead to modulate neuronal activity not only in the target but also in surrounding circuits ( 40 ). Transcranial focused ultrasound (tFUS) is emerging as a method that can complement the low degree of spatial focality of TMS and tDCS. We examine how brain stimulation can reduce seizures based on these three approaches.

Last but not least, inspiring from the recent development in seizure detection and classification method, we found that more efforts are needed to put into the following research direction to realize a complete, reliable, and low-cost seizure detection systems. First, we believe that the state-of-the-art seizure detection system performance is sufficient to build a robust and reliable wearable device that could be used for daily seizure monitoring and classification. Second, as the seizure signatures are detected and monitor, the recent brain-stimulation techniques can be used to reduce seizure. We also suggest different directions on how to build reliable and wearable seizure therapy systems. Lastly, we discuss how to build an integrated monitoring and stimulating seizure.

In the following, we first describe the state-of-the-art approach to capture physiological signals related to seizures in section 2 reliably. Next, we discuss recent efforts on building machine learning techniques to detect and classify seizures in section 3. In section 4, we discuss the different approaches to seizure therapy. Lastly, we summarize the overall contents of this article and provide the prospect of future research.

2. Analyzing Physiology Signals of Epileptic Seizure

Seizure detection and therapy systems generally consist of five processes: (1) signal acquisition, (2) signal processing, (3) feature extraction, (4) classification, and (5) therapy ( 5 , 6 , 24 ). The processes mentioned above are illustrated in Figure 1 . In this section, we discuss the needed signal processing steps to analyze the captured physiology signals of epileptic seizures. Upon the processed data, detection and classification algorithms could be built.

Figure 1 . Seizure detection and therapy overview.

2.1. Collecting Seizure-Related Signals

A seizure can be detected by monitoring various physiological signals from the human body through (1) EEG, (2) EMG, (3) ECG, (4) motion, and (5) audio/video recording ( 4 , 7 ). Among these physiological signals, EEG is the most popular choice because of its advantages, such as (1) the ability to capture the neural activation of the brain, (2) high temporal, and (3) spatial resolutions. However, the main limitation of traditional EEG measurement lies in its obtrusiveness and complicated setup, so it can only be performed in a controlled environment by a specialized technician. Also, some kinds of seizures like generalized onset motor seizures can be detected more clearly by measuring body movements or other physiological signals ( 18 , 41 ). Thus, researchers have developed seizure detection devices using various non-EEG signals as well as EEG signals ( 17 , 18 ). In the following discussion, we discuss how these recorded signals are used to detect seizure events by dividing into EEG and non-EEG methods.

2.1.1. Electroencephalogram (EEG)-Based Approach

EEG recording is the most common method to get the biosignals for seizure detection. It measures the electrical activity of the brain. Since epileptic seizure activities appear as abnormal signal patterns on the EEG, we can use the EEG signal variation to detect seizures. EEG signals with paroxysmal abnormality show spikes, spike-and-slow waves, and sharp waves in Figure 2A . Spikes are the primary form, and their time length is 20–70 ms. The spike-and-slow waves appear after spike-wave, and their time length is 200–500 ms. Sharp waves are similar to spike-wave, but their time length is 70–200 ms ( 5 ).

Figure 2 . EEG waveform. (A) EEG waveform with paroxysmal abnormality. (B) Sudden death in epilepsy recorded in ambulatory EEG ( 42 ).

The EEG recordings of patients with epileptic seizures show two categories of abnormal activity. Interictal has the abnormal signals recorded between epileptic seizures, and ictal is the activity recorded during an epileptic seizure ( 6 ). We focused on epileptic seizure detection and considered interictal and ictal EEG signals except postictal state to detect abnormal EEG signals. The EEG signature of an inter-ictal activity is occasional transient waveforms, while that of an ictal activity is composed of a continuous discharge of polymorphic waveforms of variable amplitude and frequency ( 11 ).

Many studies have been carried out for seizure detection using scalp EEG. Among them, we have selected and summarized some studies from past to recent which clearly explained the seizure detection procedure, as shown in Table 1 . Attaching EEG electrodes on all parts of the scalp is reasonable because there are many types of seizures, and the initial location which was generated the abnormal EEG signal is different. However, it causes mobility impairment, increases the cost of the measuring device, and is inappropriate for patients who need continuous seizure monitoring.

Table 1 . Seizure detection depending on the signal types.

Capturing EEG signals around the ear is a promising finding which can minimize the obtrusiveness of conventional EEG methods. The evoked responses from the ear-EEG are typically 10–20 dB lower in amplitude than those of traditional scalp EEG recordings while maintaining a similar signal-to-noise ratio (SNR) ( 62 ). Mikkelsen et al. ( 63 ) compared 32 conventional scalp electrodes with 12 ear electrodes. The measured signal from the ear electrodes reflects the same cortical activity as that from nearby scalp electrodes. Bleichner et al. ( 63 ) also worked for the comparison between a traditional EEG cap setup and their around-the-ear electrode array (cEEGrid). They have shown that their system can capture meaningful EEG signals such as eye-closing alpha wave, sleep spindles, and epileptic spike-wave. Gu et al. ( 24 ) utilized the cross-head and unilateral channels from the behind-the-ear EEG. Temporal waveform and frequency content during seizures from behind-the-ear EEG visually resembled those from scalp EEG. Especially, this paper provides the coherence between the behind-the-ear EEG channel and the best match-up scalp EEG channel on 12 patients like Figure 3 . McLean et al. ( 42 ) reported the sudden death in epilepsy recorded in ambulatory EEG. In Figure 2B , the seizure activity abruptly terminated, and the EEG became a flat line. The EEG variation graph of Figure 2B shows that these EEG channels may have significant patterns for detecting a seizure. Also, the EOG from LT-LC and RT-RC have similar morphology to that from Fp1-F7 and Fp2-F8, respectively.

Figure 3 . The best match-up scalp EEG channel of each behind-the-ear EEG channel on 12 patients ( 24 ).

2.1.2. ECG, EMG, Motion, Audio, and Video-Based Approach

EEG-based measurement usually implies that the sensors need to be attached to a human head for seizure defections. Also, EEG monitoring is prone to errors in interpreting complex signals of EEG and is mainly used to detect seizures from temporal lobe epilepsies ( 2 , 64 ). Therefore, researchers have developed seizure detection devices with various other methods. Among the relatively recent studies, we tried to select papers, which use other signals more actively than EEG and follow the clear seizure detection procedure, as shown in Table 1 . For example, contactless sensing devices such as mattress sensor [Emfit 2 , MP5 3 ], carry-on devices such as smartwatches or wrist devices [Cogan ( 65 ), Embrace ( 59 ), Inspyre 4 ], smart textiles (Neuronaute 5 ), and temporary tattoos ( 4 ) can be used to detect epilepsy. We found that ECG, EMG, motion, and audio/video recording approaches have been used to monitor epilepsy.

Electrocardiography (ECG) monitoring measures the electrical properties of the heart and detects heart rate (HR) and heart rate variability (HRV). Most of the generalized tonic-clonic seizures (GTCS) cause an increase in HR ( 66 ). Such events subsequently increase the risk of sudden unexpected death in epilepsy (SUDEP) ( 42 ). HRV is also useful to distinguish focal seizures with physical exercise ( 51 ). The most common pattern of HRV associated with focal onset impaired awareness seizures is an initial steep acceleration at the onset of the seizure ( 67 ). The HRV in temporal lobe seizures is different from that in psychogenic non-epileptic seizures ( 68 ). ECG can be used to detect a seizure. However, the accuracy and ability to detect a seizure early are still very limited.

Electromyography (EMG) monitoring measures electrical activity in response to a nerve's stimulation of the muscle. Motion is detected using accelerometers measuring the accelerations of objects in motion along reference axes ( 69 ). Both signals could be useful to detect generalized onset motor seizures.

For Audio/video recording , Arends et al. ( 56 ) evaluated the performance of audio-based detection of primary seizures (tonic-clonic and long generalized tonic). They adapted the sound threshold by training during the first week. Recognizable sounds over the threshold occurred in 23 of the 45 significant seizures. This result signifies the use of only audio recording has definite limitations. Ntonfo et al. ( 58 ) proposed the image processing approach to detect the focal clonic seizures of newborns, which are related to the periodic movements of parts of the body. They extracted an average luminance signal representative of the body movements from a video of a newborn. Single window processing has high sensitivity ( T P T P + F N , where TP : True Positive, FN : False Negative) and low specificity, ( T N T N + F P , where TN : True Negative, FP : False Positive), while multiple interlaced window processing has low sensitivity and high specificity. It is necessary to apply the advanced window protocol to improve performance.

2.1.3. Multimodality Sensing Approach

The multimodality sensing approach may improve sensitivity and lower false-positive alarms by combining the profits of each sensor, like sensing EMG signals for tonic seizure detection ( 12 ). We have chosen a number of multimodality sensing studies for the purpose of dealing more with studies that use other signals more actively than EEG, as shown in Table 1 . Electrodermal activity (EDA) refers to the variation of the electrical properties of the skin in response to sweat secretion ( 70 ). EDA is mainly used with other sensors to detect seizures, especially with ACM ( 59 ).

Cogan et al. ( 65 ) detected epileptic seizures using wrist-worn bio-sensors, which detect heart rate (HR), arterial oxygenation ( SpO 2 ), ACM, EDA, and temperature. They observed the seizure pattern of HR ↑ ⇒ SpO 2 ↓⇒ EDA ↑ . Using EEG and non-EEG signals together could be more appropriate to employ the seizure detection system precisely and extensively. Pauri et al. ( 61 ) applied EEG-video-audio monitoring to 12 patients with refractory focal seizures using 15-channel EEGs (video-cassettes). Greene et al. ( 71 ) combined EEG monitoring with ECG monitoring simultaneously for the robust detection of neonatal seizures.

2.2. Processing the Collected Signals

To collect signals, we can use wet electrodes or dry electrodes. Although the use of dry electrodes is suitable for continuous signal collection, we still need to rely on the conductive paste and gripping force of the earpieces to address the gap between the electrodes and the user's skin ( 72 ). Therefore, wet electrodes are used to maintain signal quality. And gold-plated copper electrodes are proper material due to the resistance to skin oil and sweat and rare skin allergy ( 73 ). Signal processing is necessary to get the clear biosignal waveform in the most significant frequency range (1–35 Hz) without signal distortion. Raw signal is influenced by noises from power-line and other equipment, and the signal is a mixture of several biological signals, including EEG, EOG, and EMG signals. Thus, we need to use filters.

2.2.1. Basic Filters

Typical four types of the basic filter have been used to get the clear biosignal: a low-pass, high-pass, band-pass, and notch (= band-stop) filter. In the United States, the notch filter is set at 60 Hz 6 because the 60 Hz power-line frequency noise from wires, light fluorescent, and other equipment can contaminate biosignal records. The high-pass filter can remove the low-frequency artifacts noise due to poor contact state of electrodes or the sweat of the patient under the electrodes. Furthermore, the median filter can reduce noise and high-frequency oscillations in signal data ( 74 ). However, these filters do not preserve all designated frequencies and cannot extract the specific biosignal among the overlapped biosignals spectrum ( 75 ). For example, EOG signals by eye movements or blinks propagate to the scalp electrodes creating noises in the recorded EEG signals.

2.2.2. Spatial Filters

The spatial filter technique, such as Independent Components Analysis (ICA), is a promising solution to solve the challenge of overlapped artifacts in EEG recording. Jung et al. ( 76 ) applied the spatial filters derived by ICA, which can separate and remove ocular artifacts from the recorded EEG signals. ICA technique, however, requires the use of multiple EEG electrodes to provide spatial information with the captured signals. In other words, ICA can decompose the independent components only when the number of data channels is more than that of signal sources ( 77 ). Also, ICA does not work when the training data set is too small ( 76 ).

The regression-based technique is a proposed solution to overcome the limitation of ICA. We can apply the regression approach to any number of EEG channels. Regression-based noise filtering has two phases. First, the calibration phase determines the transfer coefficients between other biosignal channels and EEG channels. Second, the correction phase estimates the noise component in the EEG recording ( 78 ). Due to this procedure, it is challenging to apply this filter in real-time. The coefficients should be controlled in the normal range. Once the coefficient is out of the range, it is not trivial for the calibration phase to turn them over to the normal range.

The wavelet-based technique is another denoising method that has been proposed for EEG signals. The wavelet-based technique compares each wavelet coefficient to a predetermined threshold and sets it to zero if its magnitude is less than the threshold ( 78 ). This technique can work in real-time and does not require the prior data of the artifacts ( 79 ). However, choosing the threshold level is a complicated process.

Lastly, if EEG recordings have multi-channel, they give blurred images of brain activity due to the volume conduction ( 80 ). In this situation, spatial filters can improve the SNR using the common spatial pattern (CSP) algorithm. The CSP extracts new time series whose variances are optimal for the discrimination of two populations of EEG based on the simultaneous diagonalization of two covariance matrices ( 81 ). Several related works have demonstrated the performance of spatial filters for multi-channel EEG ( 80 , 81 ).

2.2.3. Adaptive Filters

Adaptive filter adapts the coefficients of the filter to generate a signal similar to the noise ( 75 ). A linear adaptive filter is made up of a primary signal (= corrupted signal) d ( n ), a secondary signal (= reference signal) x ( n ), an adjustable filter H ( z ), an output from the adjustable filter y ( n ), and an error e ( n ) in Figure 4 ( 82 ). The adaptive filter usually adjusts the coefficients of filter to minimize the squared error between d ( n ) and y ( n ) ( 83 ). Correa et al. ( 84 ) arranged a cascade of three adaptive filters to remove multiple artifacts and got the EEG signal from EEG + artifacts (EOG, ECG, and power-line frequencies). However, the linear adaptive filter cannot deal with non-linear signals.

Figure 4 . Structure of an Linear Adaptive Filter.

Researchers have developed a neural network (NN) and fuzzy network (FN) to control non-linear signals. NN is made up of an input layer, hidden layer, and output layer, and users do not know the hidden layers. Fuzzy logic analyzes analog data as logical variables having continuous values between 0 and 1 ( 85 ). Each method has the following limitation. The structure of NN is challenging to decide, and the learning efficiency of FN is lower than that of NN ( 86 ). Combining them as a fuzzy neural network (FNN) is one solution to complement each drawback. However, FNN requires the training data in advance for the backpropagation, making the real-time application difficult ( 87 ).

3. Classifying and Detecting Epileptic Seizure

Seizure classification mainly categorizes the input data into one of two groups: seizure and non-seizure. Under specific requirements, the group of seizures can break down into sub-categories depending on the location of the source and symptoms. Those are considered as multiclass classification. Feature-based approaches, including feature extraction and conventional machine learning techniques, have been widely adopted to identify epileptic seizures ( 7 ). For each specific data set, the studies listed in Table 2 imposes different classifier configurations and features. Although we tried to cover recent studies in Table 2 , we also introduced some previously published papers to represent typical classification methods or data that were used well in the past. After the recent success of Deep Learning, many researchers start applying Deep Learning for medical problems, especially epileptic seizure detection/classification ( 33 ).

Table 2 . Classification for seizure detection [recommended comprehensive analysis: ( 88 , 89 )].

3.1. Feature-Based Design

In this protocol, the number and type of features have a significant impact on seizure detection performance. There are several feature extraction methods, including time-domain features, frequency-domain features, time and frequency features (discretely), and time-frequency domain features (simultaneously).

3.1.1. Feature Extraction

3.1.1.1. time-domain analysis.

Time-domain analysis works for the stationary signals, but biosignals are non-stationary. One method to quantify a non-stationary time series is to consider it as a large number of stationary segments ( 99 ). There are 12 key features in three categories: (1) mean and standard deviation for a time series with symmetric distribution; (2) median, mode, range, first quartile , and third quartile to measure the locations of a time series; (3) maximum, minimum, variation, skewness, kurtosis to pull out the shape characteristics of a time series ( 99 ). Besides, existing works have used slope sign change, Willison amplitude ( 100 ), Lyapunov exponents ( 13 ), and Hjorth parameters ( 19 ) to extract features from EEG signals.

3.1.1.2. Frequency-domain analysis

There are three basic techniques for frequency-domain analysis: Fast Fourier transform (FFT), Eigenvector, and Autoregressive ( 101 ). FFT decomposes a function (signal) of time into a frequency component fast by rearranging the input elements in a bit-reversed order and building the decimation in time ( 102 ). Fourier transform is only suitable when we are interested in what frequency components exist, not times occurring the frequency components ( 23 ). However, the time that a specific frequency component happens is essential to analyze biosignals. To solve this problem, a short-time Fourier transform (STFT) uses the idea that some part of a non-stationary signal at any given interval of time is a stationary signal. Johnson et al. ( 103 ) extracted relative power spectral density (PSD) value for each 1 Hz bin from EEG 1–40 Hz to check the state of drowsiness.

Eigenvectors are employed to calculate the signal's frequency and power from artifact dominated measurements ( 101 ). These methods are based on an eigen decomposition of the correlation matrix of the noise-corrupted signal and produce high-resolution frequency spectra even when the SNR is low ( 14 ). There are three eigenvector methods with higher resolution (Pisarenko, MUSIC, and Minimum-Norm) ( 104 ). The Pisarenko algorithm is particularly useful for estimating a spectrum containing sharp peaks at the expected frequencies ( 105 ). The MUSIC method eliminates the effects of spurious zeros by using the averaged spectra of all of the eigenvectors corresponding to the noise subspace ( 106 ). The Minimum-Norm method puts false zeros inside the unit circle and calculates the desired noise subspace vector from the eigenvectors ( 107 ).

Autoregressive methods estimate the PSD of the EEG signal using a parametric approach. These methods solve the spectral leakage problem and yield better frequency resolution ( 101 ). Yule-Walker method may lead to incorrect parameter estimates in the case of nearly periodic signals ( 108 ). As an alternative, Burg's method first estimates the reflection coefficients, and then the parameter estimates are determined using the Levinson-Durbin algorithm ( 108 ).

3.1.1.3. Time and frequency features

Using both time- and frequency-domain features can improve seizure classification performance. Srinivasan et al. ( 20 ) used three frequency-domain features (dominant frequency, average power in the primary energy zone, and normalized spectral entropy) and two time-domain features (spike rhythmicity and relative spike amplitude). Iscan et al. ( 109 ) combined time and frequency features to distinguish between seizure and healthy EEG segments. They got time-domain features using the cross-correlation method and frequency-domain features calculating the PSD.

3.1.1.4. Time-frequency domain analysis

Time-frequency domain analysis studies a signal in both the time and frequency-domains simultaneously. Time-frequency distribution (TFD) and wavelet transform analysis (WT) are the principal techniques of time-frequency domain analysis.

The basic idea of TFD is to devise a joint distribution of time and frequency that describes the energy density or intensity of a signal simultaneously in time and frequency ( 110 ). In this distribution, we can calculate the fraction of energy in a specific frequency and time range, and the distribution of frequency at a particular time. It is done by constructing a joint time-frequency function with the desired attributes and then obtaining the signal that produces the distribution ( 110 ). Boashash et al. ( 111 ) performed TFD feature extraction on multi-channel recordings for seizure detection in newborn EEG signals. They selected the optimal subset of TFD features using the wrapper method with sequential forward feature selection.

WT is an alternative to STFT. STFT gives information about the spectral components at any given interval of time, but not at a specific time instant ( 23 ). It causes a problem of resolution. WT gives a variable resolution using the characteristics that high frequencies are better resolved in time-domain, and low frequencies are in frequency-domain ( 23 ). WT can capture very minute details, sudden changes, and similarities in the EEG signals ( 22 ). It is more effective than other methods because biosignals are non-stationary ( 112 ). WT transforms a small wave (a mother wavelet) as a pattern and expresses an arbitrary waveform on the scale of magnification and reduction. WT classified into continuous wavelet transform (CWT) and discrete wavelet transform (DWT). The CWT Y ( j, k ) is defined by the following equation for any fixed-function Ψ j, k ( t ) in Equation (1). The mother wavelet (Ψ j, k ( t )) is shifted by a small interval of j in the x-axis, and correlation coefficients are computed. This procedure is repeated for various scaling factors k (dilations) in the y-axis ( 22 ).

CWT is computed by changing the scale of the analysis window, shifting the window in time, multiplying by the signal, and integrating overall times ( 23 ). However, the CWT has disadvantages such as severe redundancy of coefficients, difficulty in managing infinite wavelets, and lack of analytical methods that can easily calculate for most functions. The DWT solves these disadvantages by scaling and moving discretely, not continuously. The DWT employs two sets of functions called scaling functions and wavelet functions. These functions are related to the low-pass filter [ g ( n ), the mirror vision] and high-pass filter [ h ( n ), the discrete mother wavelet], respectively ( 113 ). In the sub-band decomposition of DWT, each stage consists of two digital filters and two downsamplers by 2. The first stage receives a signal x ( n ) and provides the detail D 1 and the approximation A 1 ( 114 ). The first approximation is further decomposed continuously. Many related works have used WT to extract features ( 115 ).

3.1.1.5. Integrating sensing signals from multiple channels

Shoeb et al. ( 116 ) extracted four features ( m = 4) representing waveform morphology on each of 21 channels ( n = 21). Then they assembled these features into a feature vector by concatenating them orderly called the early integration (EI) architecture. They also studied the performance of a patient-specific detector with an alternative architecture called the late integration (LI) architecture. In this structure, the m features of each channel assembled into a distinct feature vector and are assigned to the individual class (seizure or non-seizure). LI allows for the independent classification of activity on each channel, whereas EI summarizes interrelations between channels.

3.1.1.6. Lesson learned

Recording EEG signals is crucial because almost all seizures start from the brain. However, EEG measurement requires attaching many electrodes on the scalp with mobility impairment and making continuous measurement difficult. Therefore, we look forward to developing the devices which measure EEG signals without causing discomfort. Recording EEG signals around the ear is an emerging method to record EEG signals on the scalp. We confirmed some possibilities by comparing the similarity between EEG signal measurements around the ear and those on the scalp from many related works. Furthermore, we can get various biosignals as well as EEG around the ear ( 117 ).

The future seizure detection system is necessary to improve the signal processing procedure using spatial and adaptive filters. Basic filters do not entirely remove noise and not preserve all designated frequencies and cannot extract the specific biosignal among the overlapped biosignals spectrum ( 75 ). As we saw in Figure 3 , EEG recordings have multi-channels even around the ear. Spatial filters can improve the SNR among several channels and surrounding noises using the CSP algorithm. Adaptive filters reflect the previous signal error through the self-developed adaptive algorithm. Linear or non-linear signals are controlled depending on the adaptive algorithm.

Recent papers have applied WT to the seizure detection system to analyze the processed signals. Time-domain analysis and frequency-domain analysis are easy to use and give clearly defined features. However, they may not catch the minute features for seizure because biosignals are non-stationary. Even though there are alternatives like making a large number of stationary segments from any given interval of time, they still have a problem of resolution. Meanwhile, WT can capture very minute details, sudden changes, and similarities in the EEG signals ( 22 ). WT classified into CWT and DWT. CWT has disadvantages about the redundancy of coefficients, difficulty in managing infinite wavelets, and lack of analytical methods. DWT is usually used to solve these problems. Daubechies wavelet is the most commonly used wavelet for DWT, and the interested reader for the wavelet-based EEG processing can refer to ( 22 ) for more details.

3.1.2. Feature Classification Algorithms

3.1.2.1. support vector machine (svm).

SVM is a linear classifier that uses a hyperplane ( 25 ) to separate the data space. The mathematical expression of a hyperplane is the general form of a linear equation in multi-dimensional space.

which must have at least one a i other than zero. Given a dataset, there could be many hyperplanes that separate the data. SVM maximizes the distance between the nearest points from each group toward the hyperplane, as described in Figure 5A . This distance is called a margin. Conventional linear SVM has a limitation due to the non-linear changes of biosignals. A non-linear SVM classifier using an RBF kernel is potentially a proper approach because the seizure and non-seizure classes are not linearly separable. This approach detected 96% of 173 test seizures in a median detection delay of 3 s ( 26 ). When the categories of seizure are divided into more than two groups (e.g., focal seizure, generalized seizure, healthy), the binary classification is not sufficient to distinguish data. In this case, the SVM method for dealing with multiclass is applied to handle the problem.

Figure 5. (A) SVM (O: positive cases, X: negative cases) ( 25 ) and (B) MLPNN ( 113 ) architectures.

The development of multiclass SVM follows two approaches. One vs. rest approach is a method of binarizing the i-th class and the remaining M −1 classes. This process is repeated in the same operation for the other classes. A total of M hyperplanes are created. On the other hand, one vs. one approach is to select two of the M classes to create a hyperplane, then select the other two class combinations and repeat the same operation. A total of M ( M −1)/2 hyperplanes are created. The one vs. rest approach has an imbalance in the size of the two sets, unlike the one vs. one approach. However, the one vs. rest approach is mainly used because the total number of hyperplanes increases linearly with the number of classes. Many related works have used multiclass SVM to classify seizure states ( 29 , 30 ).

3.1.2.2. Multilayer perceptron neural network (MLPNN)

In MLPNNs , each neuron in the hidden layer sums its input signals after multiplying them by each link weights and computes its output as an activation function of the sum, as shown in Figure 5B ( 113 ). The activation function can be the rectified linear unit (ReLU), hyperbolic tangent, and so on. Guo et al. ( 92 ) used Bayesian regularization back-propagation to train MLPNN, which updates the weights and biases depending on Levenberg-Marquardt optimization. It minimizes a combination of squared errors and weights and then determines the correct combination to produce a network that generalizes well. Their network structure has one input layer with five neurons, one hidden layer with 10 neurons, and one output layer with one neuron (0—the normal/non-seizure EEG, 1—the seizure EEG). Naghsh and Aghashahi imported the feature vectors into an MLPNN system to classify the signal into three states of normal (healthy), a seizure-free interval (interictal), and a full seizure interval (ictal) ( 94 ).

3.1.2.3. Adaptive neuro-fuzzy inference system (ANFIS)

Neuro-fuzzy systems utilize the mathematical properties of ANNs in tuning rule-based fuzzy systems to approximate the way humans process information ( 95 ). Especially, ANFIS ( 118 ) has shown significant results in modeling non-linear functions. A type-3 ANFIS has five layers like Figure 6A . A circle and square indicate a fixed and adaptive node, respectively. In layer 1, the input values pass through the selected fuzzy membership function (μ A i and μ B i , i = 1, 2). This function could be a bell-shaped with a maximum equal to 1. Premise parameters { a i , b i , c i } in the function change during the training process. In layer 2, each simple multiplier multiplies the output values of layer 1 [ w i = μ A i ( x )μ B i ( y ), i = 1, 2]. In layer 3, each normalization function produces w ¯ i = w i w 1 + w 2 , i = 1 , 2 . In layer 4, the output values of layer 3 go into the Takagi and Sugeno's first-order function ( 120 ). Consequent parameters { p i , q i , r i } in the function are determined during the training process. Lastly, one single node computes the overall output as the summation of all incoming signals ( ∑ i w ¯ i f i = ∑ i w i f i ∑ i w i ). Güler and Übeyli executed a detailed classification between set A (healthy volunteer, eyes open), set B (healthy volunteer, eyes closed), set C (seizure-free intervals of five parents from hippocampal formation of opposite hemisphere), set D (seizure-free intervals of five patients from epileptogenic zone), and set E (epileptic seizure segments) using ANFIS and got the classification accuracy 98.68% ( 95 ).

Figure 6. (A) ANFIS ( 118 ) and (B) RBFNN ( 119 ) architectures.

3.1.2.4. Radial basis function neural network (RBFNN)

RBFNN is feed-forward like MLPNN but has only one hidden layer with a non-linear radial basis function (RBF) in Figure 6B ( 119 ). RBF is a real-valued function whose value depends only on the distance from the origin. RBFNN has the advantages of a simple topological structure, its locally tuned neurons, and fast learning compared to MLPNN. Aslan et al. ( 96 ) compared MLPNN with RBFNN. In the case of MLPNN, 18 out of 204 focal seizure samples were classified as a generalized seizure (8.8% error rate for focal seizure), and 9 out of 47 generalized seizure samples were classified as a focal seizure (19.1% error rate for generalized seizure). RBFNN, on the other hand, showed 3.4% and 10.6% error rate for focal and generalized seizures, respectively. However, RBFNN requires to set correct initial states. Therefore, many seizure classification papers have focused on MLPNN.

3.2. Non-feature Based Design

According to the symptoms of seizures, various types of signal patterns appear, and it is difficult to understand all of them with specific features. Thus, no existing hand-crafted features appear universally applicable so far ( 33 ). Deep learning methods can analyze the EEG signal and learn related characteristics automatically in a supervised learning framework ( 121 ). Although there are existing works that use the classification methods described as feature-based ( 20 , 98 ), we summarize these techniques in terms of non-feature based design.

3.2.1. Convolutional Neural Network (CNN)

CNN takes the raw image data and calculates the convolution by iterating over the input data according to the filter size specified to extract the feature of the data. The shape of output data changes depending on filter size, stride, padding, max-pooling size, and so on. The classifier can perform supervised learning by matching the output data and answer classes. CNN, with its high recognition performance in medical images ( 122 ), can be as good as an epileptologist in classifying seizures by analyzing EEG plot images as being observed by Emami et al. ( 33 ). In their work, they applied CNN to long-term EEG that included epileptic seizure states. In particular, EEG data were divided into short segments based on a given time window (ranging from 0.5–10 s) and converted into EEG plot images (224 × 224 pixels), each of which was classified by CNN as seizure or non-seizure. They used VGG-16 ( 123 ) because small size convolution filters (3 × 3) are capable of detecting small EEG waves. VGG-16 is also computationally efficient and can handle non-stationary objectives. This work is meaningful because the study is the first comprehensive attempt to evaluate EEG as plot images. However, the median true positive rate of CNN 74% is still low, so we cannot use this classifier for real patients.

3.2.2. Recurrent Neural Network (RNN)

In RNN , in which a network's output state depends on an arbitrary number of previous inputs like Figure 7A . However, RNN has not been widely used in applications due to the lack of an efficient and universal training method ( 124 ). Other attempts have been made to overcome these limitations. Srinivasan et al. ( 20 ) used a special type of RNN as Elman network (EN) to detect epileptic seizures. An EN has the additional set called “context layer” as shown in Figure 7B . The hidden layer is connected to these context units. Kumar et al. ( 97 ) incorporated recurrent EN and RBFNN to detect epileptic seizures with the wavelet entropy features. They showed 99.75 and 94.5% accuracy for detecting normal vs. epileptic seizures and interictal focal seizures, respectively.

Figure 7 . Recurrent neural network ( 124 ). (A) Unfolded basic. (B) Elman network.

3.3. Seizure Quantification

Biosignal quantification is necessary to make the correlations between biosignals and actual seizures more accurate ( 125 ). Adeli et al. ( 126 ) utilized the correlation dimension (CD, representing system complexity) and the largest Lyapunov exponent (LLE, representing system chaoticity) to quantify the nonlinear dynamics of the original EEGs. They analyzed three groups: group H (healthy subjects), group E (epileptic subjects during a seizure-free interval), and group S (epileptic subjects during the seizure). For the CD values from the band-limited EEGs (0–60 Hz), group S (5.3) differs from the other two groups H (6.9) and E (6.7). For the LLE values, group H (0.089), group E (0.041), and group S (0.070) differ from each other. CD and LLE have shown the possibility of being used for classification. However, to the best of our knowledge, there is no concrete explanation between the biosignal and the severity of symptoms.

4. Experimental Non-invasive Anti-seizure Treatments

We have discussed the physiological signals related to seizure as well as how to use these signals to monitor and detect seizures. The next logical step is to build a system to reduce the impact of seizure. In this section, we discuss different non-invasive brain stimulation methods that can potentially be used for seizure therapy. While we try to describe the detailed specifications, working principles, advantages, and disadvantages of different brain stimulation techniques, we leave the discussions on how to design a proper seizure therapy for future works. In particular, we discuss in detail recent non-invasive brain stimulation efforts on Transcranial Magnetic Stimulation (TMS), Transcranial Direct Current Stimulation (tDCS) ( 10 ), and Transcranial Focused Ultrasound Stimulation (tFUS) ( 40 , 127 ). Since Vagus Nerve Stimulation (VNS) overlaps tDCS in terms of electrical stimulation methods and was previously introduced primarily as invasive stimulation, it was not included in the larger category. However, recently, invasive VNS therapy for drug-resistant epilepsy patients received FDA approval 7 . VNS is also a promising seizure therapy method.

4.1. Transcranial Magnetic Stimulation

Transcranial Magnetic Stimulation (TMS) uses the principle of electromagnetic induction to focus induced current in the brain, as shown in Figure 8A ( 37 ). Strong electric currents, circulating within a coil resting on the scalp, generate short and intense magnetic fields. These magnetic fields penetrate human tissue painlessly and induce electric currents that can depolarize neurons or their axons in the brain ( 38 ). TMS techniques include single-pulse TMS (spTMS), paired-pulse TMS (ppTMS), and repetitive TMS (rTMS), as shown in Figure 9 ( 131 ). In general, single- and paired-pulse TMS are used to verify brain functions, and rTMS induces changes in brain activity that can last beyond the stimulation period ( 132 ). While spTMS and ppTMS were reported to induce unexpected seizures during multiple experiments ( 133 ), rTMS is currently a better and safer approach for seizure therapy.

Figure 8 . Brain stimulation methods to reduce the seizure symptoms. (A) Transcranial magnetic stimulation ( 128 ) (B) Transcranial direct current stimulation ( 129 ) (C) Transcranial focused ultrasound stimulation ( 130 ).

Figure 9 . TMS methods ( 131 ). (A) spTMS (B) ppTMS (C) low frequency rTMS (D) high frequency rTMS.

rTMS stimulates a single scalp site repeatedly and modulates cortical excitability. Figures 9C,D show examples tested before and after an rTMS regime. It consists of a long pattern of low (1 Hz) or high (20 Hz) frequency rTMS delivered to the left hemisphere's primary motor cortex during 28 min ( 131 , 134 ). rTMS has greater effects than single-pulse TMS but also has the potential to cause seizures ( 38 ). The FDA cleared an rTMS device as a treatment to alleviate symptoms of mildly treatment-resistant depression 8 . It shows the possibility of rTMS as a treatment for relieving various symptoms.

The effect of rTMS depends on the stimulation frequency, intensity, number of trains, inter-train interval, and number of sessions ( 135 ). We exclude the stimulus location element because it is a factor that varies depending on the symptom. The number of pulses per second of rTMS trains typically ranges between 1 and 50 Hz. One hertz paradigms are commonly applied continuously for several minutes, while higher frequency paradigms are applied in a patterned fashion like Figure 10 ( 136 ). Low-frequency rTMS produces a transient reduction in cortical excitability. High-frequency rTMS produces a local increase in cortical excitability and increases in MEP size ( 137 ). Specifically, this transient reduction effect of Low-frequency rTMS occurs in the motor cortex ( 138 ) and in the occipital cortex ( 139 ). High-frequency rTMS can improve cognitive processing to the dorsolateral prefrontal cortex ( 140 ). To compare low and high-frequency rTMS, Speer et al. ( 141 ) showed that 1-Hz rTMS was associated only with decreases in absolute regional cerebral blood flow (rCBF), while twenty-Hertz rTMS over the left prefrontal cortex was associated only with increases in rCBF.

Figure 10 . rTMS protocol example.

The stimulation intensity is usually expressed as a percentage of MT. The MT is usually determined before each session by applying the TMS coil over the primary motor cortex ( 135 ). Pulse trains are the typical form to use rTMS. If TMS stimulates the brain continuously, it can increase the possibility of generating a seizure ( 142 ) and cause heating of the electrodes. Flitman et al. ( 143 ) reported the occurrence of focal to bilateral tonic-clonic seizure in one subject after three consecutive stimulated trials with 20% above MT and pulse train lasted 750 ms at 15 Hz. Dobek et al. ( 144 ) also found 25 reports of rTMS-induced seizures in their review. Therefore, we should follow the safety guidelines for rTMS ( 145 ). Based on the international workshop on the safety of rTMS in 1996, Wassermann et al. ( 146 ) introduced the guideline for the use of rTMS: at least 5 s intervals between 20 Hz trains with intensities of up to 1.1x the MEP threshold. A longer interval is necessary for the case of higher frequencies and intensities. Bae et al. investigated the risk of seizures associated with rTMS in patients with epilepsy and reported that only 4 of 280 patients experienced seizures during or after rTMS ( 147 ).

In animal studies, low-frequency rTMS (1,000 pulses at 0.5 Hz) decreased susceptibility to pentylenetetrazol-induced seizures in rats ( 148 ). Rotenberg et al. ( 149 ) suppressed seizures in rats injected with the kainic acid using EEG-guided 0.5 and 0.75 Hz rTMS, but 0.25 Hz rTMS was not effective. In human studies, 0.3 Hz low-frequency rTMS decreased interictal EEG epileptiform abnormalities in one-third of drug-resistant epilepsy patients but was not better than a placebo for seizure reduction ( 150 ). Instead, Cincotta et al. ( 151 ) suggested that 0.3 Hz rTMS produces a relatively long-lasting enhancement of the inhibitory mechanisms responsible for the cortical silent period. Low-frequency rTMS decreased the number of seizures in patients with focal neocortical epilepsy ( 35 ) and refractory epilepsy ( 152 ).

4.2. Transcranial Direct Current Stimulation

tDCS is one of transcranial electrical stimulation (tES) methods and applies low-amplitude direct currents via scalp electrodes and penetrate the skull to enter the brain, as shown in Figure 8B ( 37 ). The principal difference between tDCS and other tES techniques is the waveform to the brain, like Figure 11 . tDCS is the only class of neuromodulation technique that delivers a sustained direct current (DC) like Figure 11A ( 39 ). Transcranial alternating current stimulation (tACS) has a variety of stimulation with different frequencies (1–45 Hz), like Figure 11B ( 153 ). tACS enables the study of causal links between brain rhythms and specific aspects of behavior. Transcranial random noise stimulation (tRNS) follows a white-noise band-limited waveform (0.1–640 Hz) like Figure 11C ( 154 ). tRNS focuses on the link between behavior and frequency-specific noise inherent in neural processing ( 127 ). tACS and tRNS are usually used to identify or compare frequency-specific characteristics. They are not actively used as therapeutic methods to obtain actual effects compared to tDCS. Besides, tDCS is easier to use, cheaper, and more tolerable than TMS. However, tDCS is still an experimental form of brain stimulation and is not an FDA-approved treatment 9 . tDCS does not induce neuronal action potentials because static fields do not yield the rapid depolarization required to produce action potentials in neural membranes ( 155 ). Thus, it is a pure neuromodulatory intervention. tDCS could modulate cortical excitation and cortical inhibition by anodal polarity and cathodal polarity, respectively. By varying the current intensity and duration, the strength and duration of the after-effects could be controlled.

Figure 11 . Waveforms of different tES techniques. (A) tDCS waveform (B) tACS waveform (C) tRNS waveform.

4.2.1. Factors

The effect of tDCS depends on current density, stimulation duration, the orientation of the electric field (the electrodes' positions and polarity), electrode configuration (material and size), the patient, deep of the target, and intensity of the current ( 156 , 157 ). Long-lasting stimulation largely influenced the durability of after-effects to humans ( 158 ). tDCS protocols should specify electrode position and current direction because these elements cause different stimulation results. The electrodes for tDCS are usually a pair of electrodes covered by sponges filled with a contact medium such as NaCl solution or conductive cream ( 155 ). For the electrode size, although large electrodes expand the area of the excitability modification, small electrodes are better to increase tDCS focality ( 155 ).

4.2.2. Case Studies

Cathodal tDCS leads to a reduction of cortical excitability by decreasing the neuronal firing rate and inducing long-term depression (LTD) of neuronal excitability ( 159 ). In animal studies, cathodal tDCS at 100 μA for 60 min resulted in a duration of more than 2 h with an increasing threshold of focal onset seizure activity, while anodal tDCS had no significant effect on TLS in the rat ( 160 ). In human studies, cathodal tDCS may be effective to reduce seizures' frequency as shown in Table 3 . Most tDCS related works applied 1–2 mA cathodal tDCS for 20–60 min. Yook et al. ( 174 ) placed a tDCS cathode at the midpoint of P4 and T4, where the 11-year-old female seizure patient showed the abnormal EEG wave. After 2 mA cathodal tDCS for 20 min, the number of seizure occurrence and the duration of each seizure episode were reduced.

Table 3 . Experimental non-invasive neuromodulation treatments for epilepsy [Reviews for readers who want to know more about neuromodulation treatments: ( 161 – 163 )].

4.2.3. Safety

Several metrics, including current density, duration, and the charge, should be controlled carefully to prevent serious adverse effects. Bikson et al. ( 175 ) investigated the related papers for the safety of tDCS and offered the criterion of tDCS protocol: current density (6.3−13 A / m 2 ) from the animal models, and others (≤ 40 min , ≤ 4 mA , ≤ 7.2 Coulombs ) from the human trials. Under this condition, there were no reports of severe side effects across over 33, 200 sessions and 1,000 subjects ( 175 ).

4.2.4. Deep Brain Stimulation of tDCS

tDCS can only directly stimulate in cortical regions. To overcome this limitation, Grossman et al. ( 176 ) suggested a new protocol called temporal interference non-invasive brain stimulation (TI-NIBS). TI delivers multiple electric fields to the brain at frequencies which are too high to recruit neural firing. These multiple electric fields differ by a frequency within the dynamic range of neural firing. They applied TI-NIBS to a living mouse brain and demonstrated the effects of TI-NIBS by stimulating neurons in subcortical structures ( 176 ). While the current experiment has not yet been applied to humans, we believe this is one of the most potential approaches for seizure therapy in the future due to its capability of providing high spatial and temporal resolution. In addition, since the stimulation is only effective at the locations where all the beams are constructive, the beams may not harm the brain cells that are not located in the targeted areas.

4.3. Transcranial Focused Ultrasound Stimulation

tFUS is emerging as a method to improve the relatively low degree of spatial locality offered by TMS and tDCS ( 127 ). It is important because a low degree of spatial locality leads to modulating neuronal activity not only in the target but also in surrounding circuits. tFUS uses acoustic energy to stimulate the brain like Figure 8C . tFUS can both excite and suppress brain neuronal activity and has millimeter spatial resolutions ( 177 ). In 1988, Colemann and Lizzi developed the Sonocare CST-100, which is the first high intensity focused ultrasound and received the FDA pre-market approval 10 . The device was designed for the treatment of glaucoma.

4.3.1. Factors

There are some factors to control the effect of tFUS: acoustic frequencies, intensities, and modes of transmission ( 178 ). First, Ultrasound (US) has a frequency above 20 kHz. Most medical US frequency range is between 1 and 15 MHz, and therapeutic US application operates around 1 MHz or less ( 179 ). Second, therapeutic US can be divided into low power (<0.5 Wcm −2 ) or high power (>100 Wcm −2 ) depending on the acoustic intensity ( 178 ). Low power intensity US is used in physiotherapy, non-thermal actions, and so on, whereas high power intensity US is used in lithotripsy and the thermal ablation of tissue ( 180 ). The therapeutic US usually has low power intensity. Lower energy US increased the action potential, while higher energy US reduced the action potential due to the ultrasonic thermal effects. Third, the US has two modes of transmission: continuous wave (CW) and pulsed wave (PW) ( 178 ). CW stimuli are more effective than PW stimuli in eliciting responses ( 181 ). Therapeutic US is usually delivered as CW or long pulse exposures ( 180 ).

4.3.2. Current Progress

In animal studies, Manlapaz et al. ( 172 ) reported ultrasonic irradiation relieved the seizures of cats. They compared fifteen cats treated by surgical removal of the epileptogenic focus with twelve cats treated by ultrasonic irradiation. The ultrasonic approach showed less postoperative complications than those of surgery. Min et al. ( 173 ) injected pentylenetetrazol to rats to induce acute epilepsy and applied tFUS to the rat's brain twice for 3 min. Epileptic EEG signals of the rats decreased visibly after tFUS compared to the other group that did not receive any tFUS. In human studies, to the best of our knowledge, there are no existing works to handle the relationship between seizure and tFUS. Instead, we look at studies that relate the human brain to tFUS. Legon et al. ( 40 ) evaluated if tFUS is capable of modulating brain activity in the human primary somatosensory cortex. From the experiment, tFUS remarkably reduced the amplitude of somatosensory evoked potential. Lee et al. ( 182 ) reported tFUS of the primary visual cortex. The tFUS induced activation both from the sonicated brain area and from the visual or cognitive network regions. However, tFUS beam might potentially harm brain cells when it passes through them.

4.3.3. Safety

It is difficult to establish tFUS protocol for safety now because there is no enough medical data about tFUS. Although Legon et al. ( 183 ) applied low power tFUS to 120 participants who did not report any neurological impairment and reported that none of the participants experienced serious adverse effects, it does not prove the safety of tFUS. This is because ultrasound at high intensities can cause irreversible tissue damage ( 40 ). The safety protocol could be established from a variety of tFUS related experiments. US studies are necessary to be conducted on primate brains such as monkeys having a skull with similar thickness and size to that of humans ( 184 ).

5. Potential Research Directions

We have discussed state-of-the-arts sensing and stimulation technologies that are suitable for seizure monitoring and therapy. In this section, we present potential research directions that require more attention to building a robust, wearable, safe, sensing, and stimulation systems.

5.1. Robust and Wearable Seizure Detection System

5.1.1. monitoring seizures from the brain with high resolution.

Current technologies only allow us to monitor the whole brain. However, we envision a more robust sensing technology that could sense precisely where the seizure signal occurs on the human brain. The future sensor can be implemented as an array of electrodes to form a beam-forming receiver to capture only the brain area of interest. It will improve the performance of the sensing system by efficiently removing the interference signals from many non-related brain areas. TI-NIBS ( 176 ) design can be considered as the closest reference.

5.1.2. Improving Seizure Quantification

Most existing seizure detection systems have only focused on differentiating between seizure and non-seizure. Therefore, we could not find a concrete explanation between the biosignal and the severity of symptoms from the related works. The biggest problem is that there are no clear criteria for quantifying seizures, and it is difficult to obtain the ground truth data. Once reasonable standards are established, and researchers' consent is received, the seizure quantification will be applied to the seizure detection system quickly.

5.1.3. Making Seizure Monitoring System Become Wearable

We predict the core location of recording EEG signals will be around the ear. The reason is that the device around the ear can still acquire clear EEG signals and does not restrict the user's mobility. Dry electrodes could be applied to the ear-cover part ( 185 ) of the device to improve usability. Different biosignals, including EOG, EMG, and EDA, also could be detected with EEG signals around the ear. A headband with EEG electrodes is also used to detect seizures from the frontal lobe or other locations which could not be detected by seizure detection devices around the ear. The wearable devices can deliver biosignals to a smartphone through communication technology. The application for seizure detection on the smartphone will extract features and classify seizure types.

5.2. Safe, Accurate, and Reliable Seizure Stimulation

5.2.1. high spacial resolution stimulation.

The stimulation device needs to localize the target area of the brain and stimulate it accurately. Existing TMS and tDCS have a low degree of spatial locality. We introduced several approaches to solve this problem. Hesed coil design of TMS attaches several strips on the specific part of the head intensively with wires that induce stimulation in the desired direction ( 186 ). TI-NIBS, as an alternative of tDCS, delivers multiple electric fields to the brain at frequencies that are too high to recruit neural firing ( 176 ). In addition, new tDCS algorithms allow a better focal treatment using multi-target electrodes and smaller electrodes in High-Definition tDCS (HD-tDCS) ( 187 ). These approaches are likely to advance to deep brain stimulation and require additional studies because they are in the proposal stage.

5.2.2. Safe and Reliable Stimulation Device

Safety is the most crucial aspect of designing the stimulation system. A practical system should prioritize the safety aspect, for example, long term and short term side effects. Unlike TMS achieving some degree of safety, tDCS and tFUS are necessary to establish the safety protocol. Antal et al. ( 188 ) introduce detail information about the safety of tDCS, including long stimulation duration, montages with (multiple) small electrodes, and limiting the maximum current. Although there are a lot of works for tDCS, tDCS still only happens in the lab environment and is not an FDA-approved treatment solutions 9 . Rather, as another electrical stimulation approach, non-invasive VNS therapy for drug-resistant epilepsy patients received FDA approval 7 . In the near future, we believe the establishment of a tDCS safety protocol for humans by integrating its new experiment with existing work. Meanwhile, there are a few related works about tFUS. Many tests for tFUS are necessary before establishing a safety protocol.

5.2.3. Making Seizure Therapy System Become Wearable

We believe tDCS could be integrated with the seizure detection system, especially around the ear. tDCS applies low-amplitude direct currents via two scalp electrodes like Figure 8B . Each scalp electrode could be connected to the surrounding area of the left ear and right ear, respectively. When designing a circuit, we need to consider the difference between the battery used for the existing seizure detection device and that used for tDCS. Unless a new design is available, it seems difficult to create a wearable device that incorporates TMS or tFUS and a seizure detection device. The fact that both brain stimulation methods are applied with a small gap between the brain and the device makes it difficult to make a wearable device.

5.3. An Integrated Sensing and Stimulation System

Even when the seizure monitoring and stimulation systems are reliable, there are many challenges remaining in integrating these two components to produce a reliable integrated system in a wearable form.

6. Conclusion

. In this paper, we systematically categorized recent efforts on building seizure monitoring, detection, and therapy systems. We explained the overall systems and components that can be used to monitor the reliable physiological signals of seizures. We presented different techniques for extracting physiological seizure signals from the noises. Then, we discussed in detail recent effort on classifying/detecting seizure events using machine learning and deep learning. Next, we presented different seizure therapy techniques, including TMS, tDCS, and tFUS. Last but not least, we discussed potential future research directions on building a wearable seizure detection and therapy system based on our experience in building comprehensive health solutions.

Author Contributions

TK surveyed and wrote all parts of the manuscript. TK and PN discussed and wrote Lesson Learned and section 6. TK, PN, NP, NB, HT, SH, and TV revised the manuscript.

Conflict of Interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

1. ^ https://www.epilepsysociety.org.uk/epileptic-seizures#.XsfrI2hKiHs

2. ^ https://www.safetysystemsdistribution.co.uk/emfit-tonic-clonic-seizure-monitor-basic/

3. ^ https://medpage-ltd.com/epileptic-tonic-clonic-seizure-alarm-MP5

4. ^ https://smart-monitor.com/about-smartwatch-inspyre-by-smart-monitor/

5. ^ https://www.bioserenity.com

6. ^ https://www.iec.ch/worldplugs/list_bylocation.htm

7. ^ https://www.mobihealthnews.com/content/fda-approves-sentiva-nerve-stimulation-device-epilepsy-therapy

8. ^ https://www.fisherwallace.com/

9. ^ https://www.hopkinsmedicine.org

10. ^ https://www.fusfoundation.org/the-technology/timeline-of-focused-ultrasound

1. Fisher RS, Boas WVE, Blume W, Elger C, Genton P, Lee P, et al. Epileptic seizures and epilepsy: definitions proposed by the International League Against Epilepsy (ILAE) and the International Bureau for Epilepsy (IBE). Epilepsia . (2005) 46:470–2. doi: 10.1111/j.0013-9580.2005.66104.x

PubMed Abstract | CrossRef Full Text | Google Scholar

2. NINDS. The Epilepsies and Seizures: Hope through Research . Department of Health & Human Services, NIH, National Institute of Health (2015).

3. Mormann F, Andrzejak RG, Elger CE, Lehnertz K. Seizure prediction: the long and winding road. Brain . (2006) 130:314–33. doi: 10.1093/brain/awl241

4. Van de Vel A, Cuppens K, Bonroy B, Milosevic M, Jansen K, Van Huffel S, et al. Non-EEG seizure detection systems and potential SUDEP prevention: state of the art: review and update. Seizure . (2016) 41:141–53. doi: 10.1016/j.seizure.2016.07.012

5. Song Y. A review of developments of EEG-based automatic medical support systems for epilepsy diagnosis and seizure detection. J Biomed Sci Eng . (2011) 4:788. doi: 10.4236/jbise.2011.412097

CrossRef Full Text | Google Scholar

6. Tzallas AT, Tsipouras MG, Tsalikakis DG, Karvounis EC, Astrakas L, Konitsiotis S, et al. Automated epileptic seizure detection methods: a review study. In: Epilepsy-Histological, Electroencephalographic and Psychological Aspects . Ioannia: InTech (2012). p. 75–98.

Google Scholar

7. Ramgopal S, Thome-Souza S, Jackson M, Kadish NE, Fernández IS, Klehm J. Seizure detection, seizure prediction, and closed-loop warning systems in epilepsy. Epilepsy Behav . (2014) 37:291–307. doi: 10.1016/j.yebeh.2014.06.023

8. Gadhoumi K, Lina JM, Mormann F, Gotman J. Seizure prediction for therapeutic devices: a review. J Neurosci Methods . (2016) 260:270–82. doi: 10.1016/j.jneumeth.2015.06.010

9. Theodore WH, Fisher RS. Brain stimulation for epilepsy. Lancet Neurol . (2004) 3:111–8. doi: 10.1016/S1474-4422(03)00664-1

10. Nitsche MA, Paulus W. Noninvasive brain stimulation protocols in the treatment of epilepsy: current state and perspectives. Neurotherapeutics . (2009) 6:244–50. doi: 10.1016/j.nurt.2009.01.003

11. McGrogan N. Neural Network Detection of Epileptic Seizures in the Electroencephalogra . Department of Engineering Science Probationary Research Transfer Report, Oxford University (1999).

12. Leijten FS, Consortium DT, van Andel J, Ungureanu C, Arends J, Tan F, et al. Multimodal seizure detection: a review. Epilepsia . (2018) 59:42–7. doi: 10.1111/epi.14047

13. Übeyli ED. Lyapunov exponents/probabilistic neural networks for analysis of EEG signals. Expert Syst Appl . (2010) 37:985–92. doi: 10.1016/j.eswa.2009.05.078

14. Übeyli ED. Analysis of EEG signals by implementing eigenvector methods/recurrent neural networks. Digital Signal Process . (2009) 19:134–43. doi: 10.1016/j.dsp.2008.07.007

15. Baumgartner C, Koren JP. Seizure detection using scalp-EEG. Epilepsia . (2018) 59:14–22. doi: 10.1111/epi.14052

16. Regalia G, Onorati F, Migliorini M, Picard R. An improved wrist-worn convulsive seizure detector based on accelerometry and electrodermal activity sensors. In: American Epilepsy Society Annual Meeting . (2015).

PubMed Abstract | Google Scholar

17. Cogan D, Birjandtalab J, Nourani M, Harvey J, Nagaraddi V. Multi-biosignal analysis for epileptic seizure monitoring. Int J Neural Syst . (2017) 27:1650031. doi: 10.1142/S0129065716500313

18. Larsen SN, Conradsen I, Beniczky S, Sorensen HB. Detection of tonic epileptic seizures based on surface electromyography. In: Engineering in Medicine Biology Society (EMBC), 36th Annual International Conference of the IEEE. Chicago, IL: IEEE (2014). p. 942–5. doi: 10.1109/EMBC.2014.6943747

19. Vidaurre C, Krämer N, Blankertz B, Schlögl A. Time domain parameters as a feature for EEG-based brain-computer interfaces. Neural Netw . (2009) 22:1313–9. doi: 10.1016/j.neunet.2009.07.020

20. Srinivasan V, Eswaran C, Sriraam N. Artificial neural network based epileptic detection using time-domain and frequency-domain features. J Med Syst . (2005) 29:647–60. doi: 10.1007/s10916-005-6133-1

21. Hassanpour H, Mesbah M, Boashash B. Time-frequency feature extraction of newborn EEG seizure using SVD-based techniques. EURASIP J Appl Signal Process . (2004) 2004:2544–54. doi: 10.1155/S1110865704406167

22. Faust O, Acharya UR, Adeli H, Adeli A. Wavelet-based EEG processing for computer-aided seizure detection and epilepsy diagnosis. Seizure . (2015) 26:56–64. doi: 10.1016/j.seizure.2015.01.012

23. Polikar R. The Wavelet Tutorial . Rowan University (1996).

24. Gu Y, Cleeren E, Dan J, Claes K, Van Paesschen W, Van Huffel S, et al. Comparison between scalp EEG and behind-the-ear EEG for development of a wearable seizure detection system for patients with focal epilepsy. Sensors . (2018) 18:29. doi: 10.3390/s18010029

25. Cortes C, Vapnik V. Support-vector networks. Mach Learn . (1995) 20:273–7. doi: 10.1007/BF00994018

26. Shoeb AH, Guttag JV. Application of machine learning to epileptic seizure detection. In: Proceedings of the 27th International Conference on Machine Learning (ICML-10) . Haifa (2010). p. 975–82.

27. Tzimourta K, Tzallas A, Giannakeas N, Astrakas L, Tsalikakis D, Tsipouras M. Epileptic seizures classification based on long-term EEG signal wavelet analysis. In: Maglaveras N, Chouvarda I, de Carvalho P, editors. Precision Medicine Powered by pHealth and Connected Health. Thessaloniki: Springer (2018). p. 165–9. doi: 10.1007/978-981-10-7419-6_28

28. Subasi A, Kevric J, Canbaz MA. Epileptic seizure detection using hybrid machine learning methods. Neural Comput. Appl . (2019) 31:317–25. doi: 10.1007/s00521-017-3003-y

29. Übeyli ED. Analysis of EEG signals by combining eigenvector methods and multiclass support vector machines. Comput. Biol. Med . (2008) 38:14–22. doi: 10.1016/j.compbiomed.2007.06.002

30. Guler I, Ubeyli ED. Multiclass support vector machines for EEG-signals classification. IEEE Trans Inform Technol Biomed . (2007) 11:117–26. doi: 10.1109/TITB.2006.879600

31. Teixeira CA, Direito B, Bandarabadi M, Le Van Quyen M, Valderrama M, Schelter B, et al. Epileptic seizure predictors based on computational intelligence techniques: a comparative study with 278 patients. Comput Methods Programs Biomed . (2014) 114:324–36. doi: 10.1016/j.cmpb.2014.02.007

32. Direito B, Teixeira CA, Sales F, Castelo-Branco M, Dourado A. A realistic seizure prediction study based on multiclass SVM. Int J Neural Syst . (2017) 27:1750006. doi: 10.1142/S012906571750006X

33. Emami A, Kunii N, Matsuo T, Shinozaki T, Kawai K, Takahashi H. Seizure detection by convolutional neural network-based analysis of scalp electroencephalography plot images. Neuroimage Clin . (2019) 22:101684. doi: 10.1016/j.nicl.2019.101684

34. Orhan U, Hekim M, Ozer M. EEG signals classification using the K-means clustering and a multilayer perceptron neural network model. Expert Syst. Appl . (2011) 38:13475–81. doi: 10.1016/j.eswa.2011.04.149

35. Santiago-Rodríguez E, Cárdenas-Morales L, Harmony T, Fernández-Bouzas A, Porras-Kattz E, Hernández A. Repetitive transcranial magnetic stimulation decreases the number of seizures in patients with focal neocortical epilepsy. Seizure . (2008) 17:677–83. doi: 10.1016/j.seizure.2008.04.005

36. Assenza G, Campana C, Formica D, Schena E, Taffoni F, Di Pino G, et al. Efficacy of cathodal transcranial direct current stimulation in drug-resistant epilepsy: a proof of principle. In: Engineering in Medicine and Biology Society (EMBC). 2014 36th Annual International Conference of the IEEE . Chicago, IL: IEEE (2014). p. 530–3. doi: 10.1109/EMBC.2014.6943645

37. Wagner T, Valero-Cabre A, Pascual-Leone A. Noninvasive human brain stimulation. Annu Rev Biomed Eng . (2007) 9:527–65. doi: 10.1146/annurev.bioeng.9.061206.133100

38. Hallett M. Transcranial magnetic stimulation and the human brain. Nature. (2000) 406:147–50. doi: 10.1038/35018000

39. Gebodh N, Esmaeilpour Z, Adair D, Schestattsky P, Fregni F, Bikson M. Transcranial direct current stimulation among technologies for low-intensity transcranial electrical stimulation: classification, history, and terminology. In: Knotkova H, Nitsche MA, Bikson M,Woods AJ, editors. Practical Guide to Transcranial Direct Current Stimulation. Cham: Springer. (2019). p. 3–43. doi: 10.1007/978-3-319-95948-1_1

40. Legon W, Sato TF, Opitz A, Mueller J, Barbour A, Williams A, et al. Transcranial focused ultrasound modulates the activity of primary somatosensory cortex in humans. Nat Neurosci . (2014) 17:322. doi: 10.1038/nn.3620

41. Nijsen TM, Aarts RM, Arends JB, Cluitmans PJ. Automated detection of tonic seizures using 3-D accelerometry. In: International Federation for Medical and Biological Engineering. Antwerp: Springer (2009) p. 188–91. doi: 10.1007/978-3-540-89208-3_47

42. McLean B, Wimalaratna S. Sudden death in epilepsy recorded in ambulatory EEG. J Neurol Neurosurg Psychiatry . (2007) 78:1395–7. doi: 10.1136/jnnp.2006.088492

43. Fisher RS, Cross JH, French JA, Higurashi N, Hirsch E, Jansen FE, et al. Operational classification of seizure types by the International League Against Epilepsy: Position Paper of the ILAE Commission for Classification and Terminology. Epilepsia . (2017) 58:522–30. doi: 10.1111/epi.13670

44. Saab M, Gotman J. A system to detect the onset of epileptic seizures in scalp EEG. Clin Neurophysiol . (2005) 116:427–42. doi: 10.1016/j.clinph.2004.08.004

45. Zandi AS, Javidan M, Dumont GA, Tafreshi R. Automated real-time epileptic seizure detection in scalp EEG recordings using an algorithm based on wavelet packet transform. IEEE Trans Biomed Eng . (2010) 57:1639–51. doi: 10.1109/TBME.2010.2046417

46. Orosco L, Correa AG, Diez P, Laciar E. Patient non-specific algorithm for seizures detection in scalp EEG. Comput Biol Med . (2016) 71:128–34. doi: 10.1016/j.compbiomed.2016.02.016

47. Wang Y, Long X, van Dijk H, Aarts RM, Wang L, Arends JB. False alarms reduction in non-convulsive status epilepticus detection via continuous EEG analysis. Physiol Meas . (2020) doi: 10.1088/1361-6579/ab8cb3

48. Doyle O, Temko A, Marnane W, Lightbody G, Boylan G. Heart rate based automatic seizure detection in the newborn. Med Eng Phys . (2010) 32:829–39. doi: 10.1016/j.medengphy.2010.05.010

49. Massé F, Bussel MV, Serteyn A, Arends J, Penders J. Miniaturized wireless ECG monitor for real-time detection of epileptic seizures. ACM Trans Embed Comput Syst . (2013) 12:102. doi: 10.1145/2485984.2485990

50. Varon C, Jansen K, Lagae L, Van Huffel S. Can ECG monitoring identify seizures? J Electrocardiol . (2015) 48:1069–74. doi: 10.1016/j.jelectrocard.2015.08.020

51. Jeppesen J, Fuglsang-Frederiksen A, Johansen P, Christensen J, Wüstenhagen S, Tankisi H, et al. Seizure detection based on heart rate variability using a wearable electrocardiography device. Epilepsia . (2019) 60:2105–13. doi: 10.1111/epi.16343

52. Vandecasteele K, De Cooman T, Gu Y, Cleeren E, Claes K, Paesschen WV, et al. Automated epileptic seizure detection based on wearable ECG and PPG in a hospital environment. Sensors . (2017) 17:2338. doi: 10.3390/s17102338

53. Szabó CÁ, Morgan LC, Karkar KM, Leary LD, Lie OV, Girouard M, et al. Electromyography-based seizure detector: Preliminary results comparing a generalized tonic-clonic seizure detection algorithm to video-EEG recordings. Epilepsia . (2015) 56:1432–7. doi: 10.1111/epi.13083

54. Van de Vel A, Cuppens K, Bonroy B, Milosevic M, Van Huffel S, Vanrumste B, et al. Long-term home monitoring of hypermotor seizures by patient-worn accelerometers. Epilepsy Behav . (2013) 26:118–25. doi: 10.1016/j.yebeh.2012.10.006

55. Beniczky S, Polster T, Kjaer TW, Hjalgrim H. Detection of generalized tonic-clonic seizures by a wireless wrist accelerometer: a prospective, multicenter study. Epilepsia . (2013) 54:e58–61. doi: 10.1111/epi.12120

56. Arends JB, van Dorp J, van Hoek D, Kramer N, van Mierlo P, van der Vorst D, et al. Diagnostic accuracy of audio-based seizure detection in patients with severe epilepsy and an intellectual disability. Epilepsy Behav . (2016) 62:180–5. doi: 10.1016/j.yebeh.2016.06.008

57. Shum J, Fogarty A, Dugan P, Holmes MG, Leeman-Markowski BA, Liu AA, et al. Sounds of seizures. Seizure . (2020) 78:86–90. doi: 10.1016/j.seizure.2020.03.008

58. Ntonfo GMK, Ferrari G, Raheli R, Pisani F. Low-complexity image processing for real-time detection of neonatal clonic seizures. IEEE Trans Inform Technol Biomed . (2012) 16:375–82. doi: 10.1109/TITB.2012.2186586

59. Poh MZ, Loddenkemper T, Reinsberger C, Swenson NC, Goyal S, Sabtala MC, et al. Convulsive seizure detection using a wrist-worn electrodermal activity and accelerometry biosensor. Epilepsia . (2012) 53:e93–7. doi: 10.1111/j.1528-1167.2012.03444.x

60. Heldberg BE, Kautz T, Leutheuser H, Hopfengärtner R, Kasper BS, Eskofier BM. Using wearable sensors for semiology-independent seizure detection-towards ambulatory monitoring of epilepsy. In: Engineering in Medicine Biology Society (EMBC), 37th Annual International Conference of the IEEE . Milan: IEEE (2015). p. 5593–6. doi: 10.1109/EMBC.2015.7319660

61. Pauri F, Pierelli F, Chatrian GE, Erdly WW. Long-term EEG-video-audio monitoring: computer detection of focal EEG seizure patterns. Electroencephalogr Clin Neurophysiol . (1992) 82:1–9. doi: 10.1016/0013-4694(92)90175-H

62. Kidmose P, Looney D, Mandic DP. Auditory evoked responses from Ear-EEG recordings. In: Engineering in Medicine and Biology Society (EMBC). 2012 Annual International Conference of the IEEE . San Diego, CA: IEEE (2012). p. 586–9. doi: 10.1109/EMBC.2012.6345999

63. Mikkelsen KB, Kappel SL, Mandic DP, Kidmose P. EEG recorded from the ear: characterizing the ear-EEG method. Front Neurosci . (2015) 9:438. doi: 10.3389/fnins.2015.00438

64. Panayiotopoulos CP. Epileptic Syndromes and their Treatment. Neonatal Seizures . 2 nd ed. London: Springer (2007). p. 185–206.

65. Cogan D, Nourani M, Harvey J, Nagaraddi V. Epileptic seizure detection using wristworn biosensors. In: Engineering in Medicine Biology Society (EMBC), 37th Annual International Conference of the IEEE . IEEE (2015). p. 5086–9. doi: 10.1109/EMBC.2015.7319535

66. Blumhardt L, Smith P, Owen L. Electrocardiographic accompaniments of temporal lobe epileptic seizures. Lancet . (1986) 327:1051–56. doi: 10.1016/S0140-6736(86)91328-0

67. Smith P, Howell S, Owen L, Blumhardt L. Profiles of instant heart rate during partial seizures. Electroencephalogr Clin Neurophysiol . (1989) 72:207–17. doi: 10.1016/0013-4694(89)90245-9

68. Ponnusamy A, Marques JL, Reuber M. Comparison of heart rate variability parameters during complex partial seizures and psychogenic nonepileptic seizures. Epilepsia . (2012) 53:1314–21. doi: 10.1111/j.1528-1167.2012.03518.x

69. Yang CC, Hsu YL. A review of accelerometry-based wearable motion detectors for physical activity monitoring. Sensors . (2010) 10:7772–88. doi: 10.3390/s100807772

70. Benedek M, Kaernbach C. A continuous measure of phasic electrodermal activity. J Neurosci Methods . (2010) 190:80–91. doi: 10.1016/j.jneumeth.2010.04.028

71. Greene BR, Boylan GB, Reilly RB, de Chazal P, Connolly S. Combination of EEG and ECG for improved automatic neonatal seizure detection. Clin Neurophysiol . (2007) 118:1348–59. doi: 10.1016/j.clinph.2007.02.015

72. Pham N, Dinh T, Raghebi Z, Kim T, Bui N, Nguyen P, et al. WAKE: a behind-the-ear wearable system for microsleep detection. In: Proceedings of the 18th Annual International Conference on Mobile Systems, Applications, and Services. Toronto, ON: ACM (2020) doi: 10.1145/3386901.3389032

73. Rapson W. Skin contact with gold and gold alloys. Contact Dermatitis . (1985) 13:56–65. doi: 10.1111/j.1600-0536.1985.tb02505.x

74. Meier R, Dittrich H, Schulze-Bonhage A, Aertsen A. Detecting epileptic seizures in long-term human EEG: a new approach to automatic online and real-time detection and classification of polymorphic seizure patterns. J Clin Neurophysiol . (2008) 25:119–31. doi: 10.1097/WNP.0b013e3181775993

75. Reddy AG, Narava S. Artifact removal from EEG signals. Int J Comput Appl. (2013) 77:17–19. doi: 10.5120/13543-1175

76. Jung TP, Makeig S, Westerfield M, Townsend J, Courchesne E, Sejnowski TJ, et al. Removal of eye activity artifacts from visual event-related potentials in normal and clinical subjects. Clin Neurophysiol . (2000) 111:1745–58. doi: 10.1016/S1388-2457(00)00386-2

77. Makeig S, Jung TP, Ghahremani D, Sejnowski TJ. Independent Component Analysis of Simulated ERP Data . Institute for Neural Computation, University of California. Technical report INC-9606. (1996).

78. Jafarifarmand A, Badamchizadeh MA. Artifacts removal in EEG signal using a new neural network enhanced adaptive filter. Neurocomputing . (2013) 103:222–31. doi: 10.1016/j.neucom.2012.09.024

79. Zikov T, Bibian S, Dumont GA, Huzmezan M, Ries C. A wavelet based de-noising technique for ocular artifact correction of the electroencephalogram. In: Engineering in Medicine and Biology. Vol. 1. Houston, TX: IEEE (2002). p. 98–105.

80. Blankertz B, Tomioka R, Lemm S, Kawanabe M, Muller KR. Optimizing spatial filters for robust EEG single-trial analysis. IEEE Signal Process Mag . (2008) 25:41–56. doi: 10.1109/MSP.2008.4408441

81. Ramoser H, Muller-Gerking J, Pfurtscheller G. Optimal spatial filtering of single trial EEG during imagined hand movement. IEEE Trans Rehabil Eng . (2000) 8:441–6. doi: 10.1109/86.895946

82. Honig ML. Adaptive Filters. Structures, Algorithms and Applications . Kluwer (1984). p. 144–246.

83. Widrow B, Stearns S. Adaptive Signal Processing . Englewood Cliffs, NJ: Prentice-Hall, Inc. (1985).

84. Correa AG, Laciar E, Patiño H, Valentinuzzi M. Artifact removal from EEG signals using adaptive filters in cascade. J Phys . (2007) 90:012081. doi: 10.1088/1742-6596/90/1/012081

85. Pedrycz W. Fuzzy Control and Fuzzy Systems . 2nd ed. Taunton, UK: Research Studies Press Ltd. (1993).

86. Chen CH, Lin CJ, Lin CT. A functional-link-based neurofuzzy network for nonlinear system control. IEEE Trans Fuzzy Syst . Taunton (2008) 16:1362–78. doi: 10.1109/TFUZZ.2008.924334

87. Er MJ, Li Z. Adaptive noise cancellation using online self-enhanced fuzzy filters with applications to multimedia processing. In: Intelligent Multimedia Processing with Soft Computing. Springer (2005) p. 389–414. doi: 10.1007/3-540-32367-8_18

88. Daoud H, Bayoumi MA. Efficient epileptic seizure prediction based on deep learning. IEEE Trans Biomed Circ Syst . (2019) 13:804–13. doi: 10.1109/TBCAS.2019.2929053

89. Craik A, He Y, Contreras-Vidal JL. Deep learning for electroencephalogram (EEG) classification tasks: a review. J Neural Eng . (2019) 16:031001. doi: 10.1088/1741-2552/ab0ab5

90. Ramakrishnan S, Murugavel AM. Epileptic seizure detection using fuzzy-rules-based sub-band specific features and layered multi-class SVM. Pattern Anal Appl . (2019) 22:1161–76. doi: 10.1007/s10044-018-0691-6

91. Andrzejak RG, Lehnertz K, Mormann F, Rieke C, David P, Elger CE. Indications of nonlinear deterministic and finite-dimensional structures in time series of brain electrical activity: dependence on recording region and brain state. Phys Rev E . (2001) 64:061907. doi: 10.1103/PhysRevE.64.061907

92. Guo L, Rivero D, Dorado J, Rabunal JR, Pazos A. Automatic epileptic seizure detection in EEGs based on line length feature and artificial neural networks. J Neurosci Methods . (2010) 191:101–9. doi: 10.1016/j.jneumeth.2010.05.020

93. Esteller R, Echauz J, Tcheng T. Comparison of line length feature before and after brain electrical stimulation in epileptic patients. In: The 26th Annual International Conference of the IEEE Engineering in Medicine and Biology Society. San Francisco, CA: IEEE (2004). p. 4710–3.

94. Naghsh-Nilchi AR, Aghashahi M. Epilepsy seizure detection using eigen-system spectral estimation and Multiple Layer Perceptron neural network. Biomed Signal Process Control . (2010) 5:147–57. doi: 10.1016/j.bspc.2010.01.004

95. Güler I, Übeyli ED. Adaptive neuro-fuzzy inference system for classification of EEG signals using wavelet coefficients. J Neurosci Methods . (2005) 148:113–21. doi: 10.1016/j.jneumeth.2005.04.013

96. Aslan K, Bozdemir H, Şahin C, Oğulata SN, Erol R. A radial basis function neural network model for classification of epilepsy using EEG signals. J Med Syst . (2008) 32:403–8. doi: 10.1007/s10916-008-9145-9

97. Kumar SP, Sriraam N, Benakop P, Jinaga B. Entropies based detection of epileptic seizures with artificial neural network classifiers. Expert Syst Appl . (2010) 37:3284–91. doi: 10.1016/j.eswa.2009.09.051

98. Hossain MS, Amin SU, Alsulaiman M, Muhammad G. Applying deep learning for epilepsy seizure detection and brain mapping visualization. ACM Trans Multimedia Comput Commun Appl . (2019) 15:10. doi: 10.1145/3241056

99. Diykh M, Li Y, Wen P. EEG sleep stages classification based on time domain features and structural graph similarity. IEEE Trans Neural Syst Rehabil Eng . (2016) 24:1159–68. doi: 10.1109/TNSRE.2016.2552539

100. Khorshidtalab A, Salami M, Hamedi M. Robust classification of motor imagery EEG signals using statistical time-domain features. Physiol Meas . (2013) 34:1563. doi: 10.1088/0967-3334/34/11/1563

101. Al-Fahoum AS, Al-Fraihat AA. Methods of EEG signal features extraction using linear analysis in frequency and time-frequency domains. ISRN Neurosci . (2014) 2014:7. doi: 10.1155/2014/730218

102. Weisstein EW. “Fourier Transform.” From MathWorld–A Wolfram Web Resource . Available online at: https://mathworld.wolfram.com/FourierTransform.html

103. Johnson RR, Popovic DP, Olmstead RE, Stikic M, Levendowski DJ, Berka C. Drowsiness/alertness algorithm development and validation using synchronized EEG and cognitive performance to individualize a generalized model. Biol Psychol . (2011) 87:241–250. doi: 10.1016/j.biopsycho.2011.03.003

104. Akay M, Semmlow JL, Welkowitz W, Bauer MD, Kostis JB. Noninvasive detection of coronary stenoses before and after angioplasty using eigenvector methods. IEEE Biomed Eng . (1990) 37:1095–104. doi: 10.1109/10.61035

105. Pisarenko VF. The retrieval of harmonics from a covariance function. Geophys J Int . (1973) 33:347–66. doi: 10.1111/j.1365-246X.1973.tb03424.x

106. Schmidt R. Multiple emitter location and signal parameter estimation. IEEE Trans Antennas Propagat . (1986) 34:276–80. doi: 10.1109/TAP.1986.1143830

107. Kumaresan R, Tufts DW. Estimating the angles of arrival of multiple plane waves. IEEE Trans Aerospace Electron Syst. (1983) AES-19:134–9. doi: 10.1109/TAES.1983.309427

108. De Hoon M, Van der Hagen T, Schoonewelle H, Van Dam H. Why Yule-Walker should not be used for autoregressive modelling. Ann Nuclear Energy . (1996) 23:1219–28. doi: 10.1016/0306-4549(95)00126-3

109. Iscan Z, Dokur Z, Demiralp T. Classification of electroencephalogram signals with combined time and frequency features. Expert Syst Appl . (2011) 38:10499–505. doi: 10.1016/j.eswa.2011.02.110

110. Cohen L. Time-frequency distributions-a review. Proc IEEE . (1989) 77:941–81. doi: 10.1109/5.30749

111. Boashash B, Ouelha S. Automatic signal abnormality detection using time-frequency features and machine learning: a newborn EEG seizure case study. Knowl Based Syst . (2016) 106:38–50. doi: 10.1016/j.knosys.2016.05.027

112. Rosso OA, Blanco S, Yordanova J, Kolev V, Figliola A, Schürmann M, et al. Wavelet entropy: a new tool for analysis of short duration brain electrical signals. J Neurosci Methods . (2001) 105:65–75. doi: 10.1016/S0165-0270(00)00356-3

113. Subasi A. EEG signal classification using wavelet feature extraction and a mixture of expert model. Expert Syst Appl . (2007) 32:1084–93. doi: 10.1016/j.eswa.2006.02.005

114. Subasi A. Epileptic seizure detection using dynamic wavelet network. Expert Syst Appl . (2005) 29:343–55. doi: 10.1016/j.eswa.2005.04.007

115. Acharya UR, Sree SV, Ang PCA, Yanti R, Suri JS. Application of non-linear and wavelet based features for the automated identification of epileptic EEG signals. Neural Syst . (2012) 22:1250002. doi: 10.1142/S0129065712500025

116. Greene B, Marnane W, Lightbody G, Reilly R, Boylan G. Classifier models and architectures for EEG-based neonatal seizure detection. Physiol Meas . (2008) 29:1157. doi: 10.1088/0967-3334/29/10/002

117. Nguyen P, Bui N, Nguyen A, Truong H, Suresh A, Whitlock M, et al. Tyth-typing on your teeth: tongue-teeth localization for human-computer interface. In: Proceedings of the 16th Annual International Conference on Mobile Systems, Applications, and Services. Munich: ACM (2018). p. 269–82. doi: 10.1145/3210240.3210322

118. Jang JS. ANFIS: adaptive-network-based fuzzy inference system. IEEE Trans Syst Man Cybernet . (1993) 23:665–85. doi: 10.1109/21.256541

119. Moody J, Darken CJ. Fast learning in networks of locally-tuned processing units. Neural Comput . (1989) 1:281–94. doi: 10.1162/neco.1989.1.2.281

120. Takagi T, Sugeno M. Derivation of fuzzy control rules from human operator's control actions. IFAC Proc Vol . (1983) 16:55–60. doi: 10.1016/S1474-6670(17)62005-6

121. LeCun Y, Bengio Y, Hinton G. Deep learning. Nature . (2015) 521:436–44. doi: 10.1038/nature14539

122. Litjens G, Kooi T, Bejnordi BE, Setio AAA, Ciompi F, Ghafoorian M, et al. A survey on deep learning in medical image analysis. Med Image Anal . (2017) 42:60–88. doi: 10.1016/j.media.2017.07.005

123. Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:14091556 (2014).

124. Petrosian A, Prokhorov D, Homan R, Dasheiff R, Wunsch D II. Recurrent neural network based prediction of epileptic seizures in intra-and extracranial EEG. Neurocomputing . (2000) 30:201–18. doi: 10.1016/S0925-2312(99)00126-5

125. Gotman J, Gloor P. Automatic recognition and quantification of interictal epileptic activity in the human scalp EEG. Electroencephalogr Clin Neurophysiol . (1976) 41:513–29. doi: 10.1016/0013-4694(76)90063-8

126. Adeli H, Ghosh-Dastidar S, Dadmehr N. A wavelet-chaos methodology for analysis of EEGs and EEG subbands to detect seizure and epilepsy. IEEE Trans Biomed Eng . (2007) 54:205–11. doi: 10.1109/TBME.2006.886855

127. Polania R, Nitsche MA, Ruff CC. Studying and modifying brain function with non-invasive brain stimulation. Nat Neurosci. (2018) 21: 174–87. doi: 10.1038/s41593-017-0054-4

128. Ridding MC, Rothwell JC. Is there a future for therapeutic use of transcranial magnetic stimulation? Nat Rev Neurosci . (2007) 8:559. doi: 10.1038/nrn2169

129. Higgins ES, George MS. Brain Stimulation Therapies for Clinicians . Sullivan's Island, SC: American Psychiatric Publication (2019).

130. Kim J, Lee S. Development of a wearable robotic positioning system for noninvasive transcranial focused ultrasound stimulation. IEEE/ASME Trans Mechatron . (2016) 21:2284–93. doi: 10.1109/TMECH.2016.2580500

131. Valero-Cabré A, Amengual JL, Stengel C, Pascual-Leone A, Coubard OA. Transcranial magnetic stimulation in basic and clinical neuroscience: a comprehensive review of fundamental principles and novel insights. Neurosci Biobehav Rev . (2017) 83:381–404. doi: 10.1016/j.neubiorev.2017.10.006

132. Klomjai W, Katz R, Lackmy-Vallée A. Basic principles of transcranial magnetic stimulation (TMS) and repetitive TMS (rTMS). Ann Phys Rehabil Med . (2015) 58:208–13. doi: 10.1016/j.rehab.2015.05.005

133. Schrader LM, Stern JM, Koski L, Nuwer MR, Engel J Jr. Seizure incidence during single-and paired-pulse transcranial magnetic stimulation in individuals with epilepsy. Clin Neurophysiol . (2004) 115:2728–37. doi: 10.1016/j.clinph.2004.06.018

134. Gangitano M, Valero-Cabré A, Tormos JM, Mottaghy FM, Romero JR, Pascual-Leone Á. Modulation of input-output curves by low and high frequency repetitive transcranial magnetic stimulation of the motor cortex. Clin Neurophysiol . (2002) 113:1249–57. doi: 10.1016/S1388-2457(02)00109-8

135. Spronk D, Arns M, Fitzgerald PB. Repetitive transcranial magnetic stimulation in depression: protocols, mechanisms, and new developments. In: Neurofeedback and Neuromodulation Techniques and Applications . Academic Press (2011). p. 257–91. doi: 10.1016/B978-0-12-382235-2.00010-X

136. Oberman L. Repetitive transcranial magnetic stimulation (rTMS) protocols. In: Rotenberg A, Horvath JC, Pascual-Leone A, editors. Transcranial Magnetic Stimulation . New York, NY: Springer (2014). p. 129–39. doi: 10.1007/978-1-4939-0879-0_7

137. Chen R, Seitz RJ. Changing cortical excitability with low-frequency magnetic stimulation. Neurology . (2001) 57:379–80. doi: 10.1212/WNL.57.3.379

138. Plewnia C, Lotze M, Gerloff C. Disinhibition of the contralateral motor cortex by low-frequency rTMS. Neuroreport . (2003) 14:609–12. doi: 10.1097/00001756-200303240-00017

139. Thut G, Theoret H, Pfennig A, Ives J, Kampmann F, Northoff G, et al. Differential effects of low-frequency rTMS at the occipital pole on visual-induced alpha desynchronization and visual-evoked potentials. Neuroimage . (2003) 18:334–47. doi: 10.1016/S1053-8119(02)00048-4

140. Rollnik JD, Huber TJ, Mogk H, Siggelkow S, Kropp S, Dengler R, et al. High frequency repetitive transcranial magnetic stimulation (rTMS) of the dorsolateral prefrontal cortex in schizophrenic patients. Neuroreport . (2000) 11:4013–5. doi: 10.1097/00001756-200012180-00022

141. Speer AM, Kimbrell TA, Wassermann EM, Repella JD, Willis MW, Herscovitch P, et al. Opposite effects of high and low frequency rTMS on regional brain activity in depressed patients. Biol Psychiatry . (2000) 48:1133–41. doi: 10.1016/S0006-3223(00)01065-9

142. Hallett M. Transcranial magnetic stimulation: a primer. Neuron . (2007) 55:187–99. doi: 10.1016/j.neuron.2007.06.026

143. Flitman S, Grafman J, Wassermann E, Cooper V, O'grady J, Pascual-Leone A, et al. Linguistic processing during repetitive transcranial magnetic stimulation. Neurology . (1998) 50:175–81. doi: 10.1212/WNL.50.1.175

144. Dobek CE, Blumberger DM, Downar J, Daskalakis ZJ, Vila-Rodriguez F. Risk of seizures in transcranial magnetic stimulation: a clinical review to inform consent process focused on bupropion. Neuropsychiatr Dis Treat . (2015) 11:2975. doi: 10.2147/NDT.S91126

145. Rossi S, Hallett M, Rossini PM, Pascual-Leone A Safety of TMS Consensus Group. Safety, ethical considerations, and application guidelines for the use of transcranial magnetic stimulation in clinical practice and research. Clin Neurophysiol . (2009) 120:2008–39. doi: 10.1016/j.clinph.2009.08.016

146. Wassermann EM. Risk and safety of repetitive transcranial magnetic stimulation: report and suggested guidelines from the International Workshop on the Safety of Repetitive Transcranial Magnetic Stimulation, June 5-7, 1996. Electroencephalogr Clin Neurophysiol Evoked Potent Sec . (1998) 108:1–16. doi: 10.1016/S0168-5597(97)00096-8

147. Bae EH, Schrader LM, Machii K, Alonso-Alonso M, Riviello JJ Jr, Pascual-Leone A, et al. Safety and tolerability of repetitive transcranial magnetic stimulation in patients with epilepsy: a review of the literature. Epilepsy Behav . (2007) 10:521–8. doi: 10.1016/j.yebeh.2007.03.004

148. Akamatsu N, Fueta Y, Endo Y, Matsunaga K, Uozumi T, Tsuji S. Decreased susceptibility to pentylenetetrazol-induced seizures after low-frequency transcranial magnetic stimulation in rats. Neurosci Lett . (2001) 310:153–6. doi: 10.1016/S0304-3940(01)02116-4

149. Rotenberg A, Muller P, Birnbaum D, Harrington M, Riviello JJ, et al. Seizure suppression by EEG-guided repetitive transcranial magnetic stimulation in the rat. Clin Neurophysiol . (2008) 119:2697–702. doi: 10.1016/j.clinph.2008.09.003

150. Cantello R, Rossi S, Varrasi C, Ulivelli M, Civardi C, Bartalini S, et al. Slow repetitive TMS for drug-resistant epilepsy: clinical and EEG findings of a placebo-controlled trial. Epilepsia . (2007) 48:366–74. doi: 10.1111/j.1528-1167.2006.00938.x

151. Cincotta M, Borgheresi A, Gambetti C, Balestrieri F, Rossi L, Zaccara G, Pascual-Leone A, et al. Suprathreshold 0.3 Hz repetitive TMS prolongs the cortical silent period: potential implications for therapeutic trials in epilepsy. Clin Neurophysiol . (2003) 114:1827–33. doi: 10.1016/S1388-2457(03)00181-0

152. Fregni F, Otachi PT, Do Valle A, Boggio PS, Thut G, Rigonatti SP, et al. A randomized clinical trial of repetitive transcranial magnetic stimulation in patients with refractory epilepsy. Ann Neurol . (2006) 60:447–55. doi: 10.1002/ana.20950

153. Herrmann CS, Rach S, Neuling T, Strüber D. Transcranial alternating current stimulation: a review of the underlying mechanisms and modulation of cognitive processes. Front Hum Neurosci . (2013) 7:279. doi: 10.3389/fnhum.2013.00279

154. Reato D, Salvador R, Bikson M, Opitz A, Dmochowski J, Miranda PC. Principles of transcranial direct current stimulation (tDCS): introduction to the biophysics of tDCS. In: Knotkova H, Nitsche MA, Bikson M, Woods AJ, editors. Practical Guide to Transcranial Direct Current Stimulation . Cham: Springer (2019). p. 45–80. doi: 10.1007/978-3-319-95948-1_2

155. Nitsche MA, Cohen LG, Wassermann EM, Priori A, Lang N, Antal A, et al. Transcranial direct current stimulation: state of the art (2008). Brain Stimul . (2008) 1:206–23. doi: 10.1016/j.brs.2008.06.004

156. Nitsche MA, Paulus W. Excitability changes induced in the human motor cortex by weak transcranial direct current stimulation. J Physiol . (2000) 527:633–9. doi: 10.1111/j.1469-7793.2000.t01-1-00633.x

157. Peterchev AV, Wagner TA, Miranda PC, Nitsche MA, Paulus W, Lisanby SH, et al. Fundamentals of transcranial electric and magnetic stimulation dose: definition, selection, and reporting practices. Brain Stimul . (2012) 5:435–53. doi: 10.1016/j.brs.2011.10.001

158. Nitsche MA, Paulus W. Sustained excitability elevations induced by transcranial DC motor cortex stimulation in humans. Neurology . (2001) 57:1899–901. doi: 10.1212/WNL.57.10.1899

159. Palm U, Segmiller FM, Epple AN, Freisleder FJ, Koutsouleris N, Schulte-Körne G, et al. Transcranial direct current stimulation in children and adolescents: a comprehensive review. J Neural Trans . (2016) 123:1219–34. doi: 10.1007/s00702-016-1572-z

160. Liebetanz D, Klinker F, Hering D, Koch R, Nitsche MA, Potschka H, et al. Anticonvulsant effects of transcranial direct-current stimulation (tDCS) in the rat cortical ramp model of focal epilepsy. Epilepsia . (2006) 47:1216–24. doi: 10.1111/j.1528-1167.2006.00539.x

161. Schulze-Bonhage A. Brain stimulation as a neuromodulatory epilepsy therapy. Seizure . (2017) 44:169–75. doi: 10.1016/j.seizure.2016.10.026

162. Larkin M, Meyer RM, Szuflita NS, Severson MA, Levine ZT. Post-traumatic, drug-resistant epilepsy and review of seizure control outcomes from blinded, randomized controlled trials of brain stimulation treatments for drug-resistant epilepsy. Cureus . (2016) 8:16. doi: 10.7759/cureus.744

163. Li MC, Cook MJ. Deep brain stimulation for drug-resistant epilepsy. Epilepsia . (2018) 59:273–90. doi: 10.1111/epi.13964

164. Tergau F, Naumann U, Paulus W, Steinhoff BJ. Low-frequency repetitive transcranial magnetic stimulation improves intractable epilepsy. Lancet . (1999) 353:2209. doi: 10.1016/S0140-6736(99)01301-X

165. Rossi S, Ulivelli M, Bartalini S, Galli R, Passero S, Battistini N, et al. Reduction of cortical myoclonus-related epileptic activity following slow-frequency rTMS. A case study. Neuroreport . (2004) 15:293–96. doi: 10.1097/00001756-200402090-00016

166. Sun W, Mao W, Meng X, Wang D, Qiao L, Tao W, et al. Low-frequency repetitive transcranial magnetic stimulation for the treatment of refractory partial epilepsy: a controlled clinical study. Epilepsia . (2012) 53:1782–9. doi: 10.1111/j.1528-1167.2012.03626.x

167. Auvichayapat N, Rotenberg A, Gersner R, Ngodklang S, Tiamkao S, Tassaneeyakul W, et al. Transcranial direct current stimulation for treatment of refractory childhood focal epilepsy. Brain Stimul . (2013) 6:696–700. doi: 10.1016/j.brs.2013.01.009

168. Tekturk P, Erdogan ET, Kurt A, Vanli-yavuz EN, Ekizoglu E, Kocagoncu E, et al. The effect of transcranial direct current stimulation on seizure frequency of patients with mesial temporal lobe epilepsy with hippocampal sclerosis. Clin Neurol Neurosurg . (2016) 149:27–32. doi: 10.1016/j.clineuro.2016.07.014

169. Assenza G, Campana C, Assenza F, Pellegrino G, Di Pino G, Fabrizio E, et al. Cathodal transcranial direct current stimulation reduces seizure frequency in adults with drug-resistant temporal lobe epilepsy: a sham controlled study. Brain Stimul Basic Transl Clin Res Neuromodul . (2017) 10:333–5. doi: 10.1016/j.brs.2016.12.005

170. San-Juan D, López DAE, Gregorio RV, Trenado C, Gonzalez-Aragon MF, Morales-Quezada L, et al. Transcranial direct current stimulation in mesial temporal lobe epilepsy and hippocampal sclerosis. Brain Stimul . (2017) 10:28–35. doi: 10.1016/j.brs.2016.08.013

171. Karvigh SA, Motamedi M, Arzani M, Roshan JHN. HD-tDCS in refractory lateral frontal lobe epilepsy patients. Seizure . (2017) 47:74–80. doi: 10.1016/j.seizure.2017.03.005

172. Manlapaz J, Åström K, Ballantine H Jr, Lele P. Effects of ultrasonic radiation in experimental focal epilepsy in the cat. Exp Neurol . (1964) 10:345–56. doi: 10.1016/0014-4886(64)90005-6

173. Min BK, Bystritsky A, Jung KI, Fischer K, Zhang Y, Maeng LS, et al. Focused ultrasound-mediated suppression of chemically-induced acute epileptic EEG activity. BMC Neurosci . (2011) 12:23. doi: 10.1186/1471-2202-12-23

174. Yook SW, Park SH, Seo JH, Kim SJ, Ko MH. Suppression of seizure by cathodal transcranial direct current stimulation in an epileptic patient-a case report. Ann Rehabil Med . (2011) 35:579–82. doi: 10.5535/arm.2011.35.4.579

175. Bikson M, Grossman P, Thomas C, Zannou AL, Jiang J, Adnan T, et al. Safety of transcranial direct current stimulation: evidence based update (2016). Brain Stimul . (2016) 9:641–61. doi: 10.1016/j.brs.2016.06.004

176. Grossman N, Bono D, Dedic N, Kodandaramaiah SB, Rudenko A, Suk HJ, et al. Noninvasive deep brain stimulation via temporally interfering electric fields. Cell . (2017) 169:1029–41. doi: 10.1016/j.cell.2017.05.024

177. Yang T, Chen J, Yan B, Zhou D. Transcranial ultrasound stimulation: a possible therapeutic approach to epilepsy. Med Hypotheses . (2011) 76:381–3. doi: 10.1016/j.mehy.2010.10.046

178. Tufail Y, Yoshihiro A, Pati S, Li MM, Tyler WJ. Ultrasonic neuromodulation by brain stimulation with transcranial ultrasound. Nat Protoc . (2011) 6:1453. doi: 10.1038/nprot.2011.371

179. O B̀rien WD Jr. Ultrasound-biophysics mechanisms. Biophys Mol Biol . (2007) 93:212–255. doi: 10.1016/j.pbiomolbio.2006.07.010

180. Ter Haar G. Therapeutic applications of ultrasound. Biophys Mol Biol . (2007) 93:111–29. doi: 10.1016/j.pbiomolbio.2006.07.005

181. King RL, Brown JR, Newsome WT, Pauly KB. Effective parameters for ultrasound-induced in vivo neurostimulation. Ultrasound Med Biol . (2013) 39:312–31. doi: 10.1016/j.ultrasmedbio.2012.09.009

182. Lee W, Kim HC, Jung Y, Chung YA, Song IU, Lee JH, et al. Transcranial focused ultrasound stimulation of human primary visual cortex. Sci Rep . (2016) 6:34026. doi: 10.1038/srep34026

PubMed Abstract | CrossRef Full Text

183. Legon W, Bansal P, Ai L, Mueller JK, Meekins G, Gillick B. Safety of transcranial focused ultrasound for human neuromodulation. bioRxiv [Preprint] . (2018) 314856. doi: 10.1101/314856

184. Rezayat E, Toostani IG. A review on brain stimulation using low intensity focused ultrasound. Basic Clin Neurosci . (2016) 7:187. doi: 10.15412/J.BCN.03070303

185. Bleichner MG, Mirkovic B, Debener S. Identifying auditory attention with ear-EEG: cEEGrid versus high-density cap-EEG comparison. J Neural Eng . (2016) 13:066004. doi: 10.1088/1741-2560/13/6/066004

186. Roth Y, Zangen A, Hallett M. A coil design for transcranial magnetic stimulation of deep brain regions. J Clin Neurophysiol . (2002) 19:361–70. doi: 10.1097/00004691-200208000-00008

187. Edwards D, Cortes M, Datta A, Minhas P, Wassermann EM, Bikson M. Physiological and modeling evidence for focal transcranial electrical brain stimulation in humans: a basis for high-definition tDCS. Neuroimage . (2013) 74:266–75. doi: 10.1016/j.neuroimage.2013.01.042

188. Antal A, Alekseichuk I, Bikson M, Brockmöller J, Brunoni AR, Chen R, et al. Low intensity transcranial electric stimulation: safety, ethical, legal regulatory and application guidelines. Clin Neurophysiol . (2017) 128:1774–809. doi: 10.1016/j.clinph.2017.06.001

Keywords: seizure detection, biosignal processing, biosignal classification, brain stimulation, EEG

Citation: Kim T, Nguyen P, Pham N, Bui N, Truong H, Ha S and Vu T (2020) Epileptic Seizure Detection and Experimental Treatment: A Review. Front. Neurol. 11:701. doi: 10.3389/fneur.2020.00701

Received: 11 February 2020; Accepted: 09 July 2020; Published: 21 July 2020.

Reviewed by:

Copyright © 2020 Kim, Nguyen, Pham, Bui, Truong, Ha and Vu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Tam Vu, tam.vu@colorado.edu ; tam.vu@cs.ox.ac.uk

Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.

• Open access
• Published: 19 January 2024

Detection of epileptic seizure in EEG signals using machine learning and deep learning techniques

• Pankaj Kunekar   ORCID: orcid.org/0000-0001-6729-5315 1 ,
• Mukesh Kumar Gupta 1 &
• Pramod Gaur 2  

Journal of Engineering and Applied Science volume  71 , Article number:  21 ( 2024 ) Cite this article

1582 Accesses

1 Citations

Metrics details

Around 50 million individuals worldwide suffer from epilepsy, a chronic, non-communicable brain disorder. Several screening methods, including electroencephalography, have been proposed to identify epileptic episodes. EEG data, which are frequently utilised to enhance epilepsy analysis, offer essential information on the electrical processes of the brain. Prior to the emergence of deep learning (DL), feature extraction was accomplished by standard machine learning techniques. As a result, they were only as good as the people who made the features by hand. But with DL, both feature extraction and classification are fully automated. These methods have significantly advanced several fields of medicine, including the diagnosis of epilepsy. In this paper, the works focused on automated epileptic seizure detection using ML and DL techniques are presented as well as their comparative analysis is done. The UCI-Epileptic Seizure Recognition dataset is used for training and validation. Some of the conventional ML and DL algorithms are used with a proposed model which uses long short-term memory (LSTM) to find the best approach. Post that comparative analysis is performed on these algorithms to find the best approach for epileptic seizure detection. As a result, the proposed model LSTM gives a validation accuracy of 97% giving the most appropriate and precise result as compared to other mentioned algorithms used in this study.

Introduction

Epilepsy is a persistent, non-communicable brain condition. A hereditary condition or an acquired brain disorder, such as a trauma or stroke, may cause epilepsy. A person who is having a seizure exhibits strange behaviour, symptoms and sensations, sometimes even losing consciousness. According to the most recent assessment by the World Health Organization (WHO), around 50 million people worldwide experience epileptic seizures, and the majority of them are unaware of their illness.

Most of the time seizure attacks are the cause of accidents. Seizures are a result of excessive electrical discharges in a group of brain cells. A patient who has a seizure attack for more than 5 min needs to be medicated as soon as possible, but due to the lack of knowledge about this condition, it is not treated at early stages. It is not possible to treat any brain-related condition just by observing the patient. The ionic current passing through the brain neurons is observed using electroencephalogram (EEG) which provides a graph of temporal and spatial information about the brain. However, it is challenging to obtain comprehensive information about these dynamic biological signals due to the non-linear and non-stationary nature of EEG signals.

Depending on the seizure characteristics, epilepsy has various types of epilepsy seizures [ 1 ]. In recent years, treatment of epileptic seizures is possible due to advancements in the medical field, but still, detection and classification of epileptic seizures are tedious tasks without using automation that can be done using various machine learning and deep learning techniques. Also, there are different times to observe patient brain signals. Using various ML and DL techniques on EEG signal data to detect seizure helps to get insights into a patient’s brain condition more preciously.

There is a lot of work done in this domain already, but most algorithm fails to get validation accuracy and other model performance parameters. Also, the implemented model sometimes fails because of a lack of dataset.

The proposed system is intended to solve the problem of automation in epileptic seizure detection. The proposed model uses LSTM to detect seizure patients or normal persons by considering observations from the UCI-Epileptic Seizure detection dataset. Also, the designed system matches the research gaps in this area by providing a more accurate classification of the signal into two categories.

The proposed model is implemented by considering various ML algorithms and deep learning algorithms and their performance on the UCI dataset, so that the validation accuracy can be increased by observing other models’ performance. The detailed comparison of machine learning and deep learning models is done with a proposed model to solve many problems.

Related work

Various papers were studied to know about the research done in the area. Some of the important surveys and techniques are explained below:

In [ 2 ], a survey on scalable technologies assisting in early screening and predictive analysis for lifestyle diseases is done. Study [ 3 ] explores various forms of artificial intelligence on large volumes of data in medical technology. In [ 4 , 5 , 6 ], various studies explored that machine learning and deep learning techniques have been efficiently used for enhanced medical innovations.

In the paper [ 7 ], the authors proposed a system which uses empirical mode decomposition (EMD) and extracted time and frequency domain features for extracting features from EEG signals. The CHB-MIT dataset is used here. The model has given a higher true-positive rate of 92.23%, and the average prediction time is 23.60 min on the scalp for all subjects in the dataset.

In the paper [ 8 ], emphasis is placed on computational complexity, and other models are also compared by systems with multiple classifications. The three main basic parts are as follows: the first one is feature extraction, the second one is hierarchical attention layer and the last is classification layer. Three scenarios were used to evaluate this system: based on combinations of interictal, preictal and ictal.

In the paper [ 9 ], the author proposed a system focused on feature extraction and classification. The model used Taylor Fourier, rhythm-specific and filter bank for pre-processing and at the end feature extraction and classification SVM used. The model achieved an accuracy of 94.88%. The Bonn University Database is used here.

In the paper [ 10 ], using a corpus of EEG data from Temple University Hospital, the authors applied CNN and transfer learning to categorise seven variations of seizures with non-seizure EEG. This study’s goal is to carry out a multiclass categorisation. Before being fed as input to CNN, the signal was transformed into a spectrogram. To choose the best network for the given study, many DL pre-trained networks were employed. 82.85% (Googlenet) and 88.30% (Inceptionv3) are the best categorisation accuracy models used in this study.

In this system [ 11 ], to examine alternative model assessment parameters, two fusion methods were considered: the first one is ensemble, and the second one is Choquet fuzzy integral used with the deep neural network used in the system.

In the paper [ 12 ], a singular value decomposition fuzzy k -nearest neighbour classifier methodology based on discrete wavelet transform offers about 100% accuracy and nearly about 93.33% on two and three classes using the Bonn University Dataset.

In [ 13 ], the authors categorise EEG data into ictal and interictal types. The primary problem here is the non-stationary and non-linear character of the EEG signal while attempting to understand brain output. The main emphasis is on feature extraction from seizure EEG recordings to create a method for epileptic seizure identification on the CHB-MIT dataset that uses both fuzzy-based and conventional machine learning techniques.

In [ 14 ], the authors suggest using deep learning to identify seizures in paediatric patients. To implement the supervised classifier, the CHB-MIT dataset is used to classify ictal and interictal brain state signals. Two-dimensional deep convolution auto-encoder connected to a neural network.

The paper [ 15 ] intended to use the principal components analysis used for the feature reduction approach to the signals in order to obtain the optimum classification algorithm for epileptic seizures. Using the dataset to predict epilepsy, KNN, RF, SVM, ANN and DT algorithms are used, and the performance of classifiers is examined both with and without PCA.

This paper [ 16 ] reviews different deep learning algorithms with CNN for 1D CNN, 2D CNN, CNN using transfer learning and LSTM. The author reviewed different approaches for each technique in which the CNN and LSTM were showing significant accuracy of around 99%. The different dataset was considered CHB-MIT, BON, Flint Hills and Bern Barcelona.

The paper [ 17 ] presents a model which solves the issues of data imbalance, low accuracy and classification model with sampling techniques including downsampling, random sampling and the synthetic minority oversampling technique. The authors proposed the heterogeneous deep ensemble model which gives an accuracy score of 0.93) and an F -measure value of 0.91.

The paper’s [ 18 ] authors say manual observation of EEG is performed to detect epilepsy which makes it difficult and easy to switch over automated diagnosis system. The Bonn EEG Dataset is used. The authors proposed a least squares support vector machine (LSSVM) as a better approach to deal with linear equations with an accuracy of 94.7%.

In the paper [ 19 ], the authors worked on EEG signal noise removal. The EEG signal gets compromised by background noise or any muscle movement which makes it difficult to detect in automatic mode. After taking this limitation, the paper reviews different automatic approaches which state feature selection and classification are the tedious and error-prone area in epilepsy.

In the paper [ 20 ], the authors proposed ResNET-50 as an automated system which will define the EEG data into non-ictal, ictal and pre-ictal classes. The CHB-MIT, Freiburg, BONN Dataset and BERN Dataset use CNN by transforming the 2D EEG images from 1D EEG images which give 94.88% accuracy.

In the paper [ 21 ], the authors proposed CNN for the classification of an epileptic seizure. The proposed model contains four models: the CNN model, fusion of two CNN; fusion of 3 CNN, fusion of 4 CNN model and transfer learning using ResNet 50. The fusion of 3 and 4 CNN models gave significantly best results with 95% accuracy. The two convolution layers with 32 filter and 3 × 3 kernel size are used as a single CNN model which after concatenated for fusion of 2, 3 and 4 CNN models.

In the paper [ 22 ], the authors proposed a deep neural network with hierarchical attention mechanisms. The system starting layer was of two separate CNN for extracting the feature and connected to the hierarchical attention layer and fully connected layers for classification. The computation time of the proposed model was 0.23 s for classification and 0.014 for feature extraction.

In the paper [ 23 ], a novel seizure prediction model called TASM ResNet was proposed by the authors. It is based on an intracranial EEG signal-based pre-trained ResNet and a temporal attention simulation module. The simulation module was created to take raw EEG data, transform it into data that resembles images and then extract temporal characteristics. ResNet was utilised in this case to decrease the amount of training data. Additionally, the final outcomes demonstrated that an image network that has been pre-trained on a sizable dataset using a simulation module can migrate EEG signals.

Proposed methods

The proposed system follows the traditional way of model training and testing as shown in Fig.  1 . The UCI-Epileptic Seizure Dataset is firstly pre-processed to feed to the model. After dataset pre-processing done, the dataset split into training and testing data. After that model is selected, there are logistic regression, KNN, SVM, ANN and LSTM. Next, the selected model is trained based on UCI training dataset. The testing data is feed to model to get results. Finally, the model was evaluated based on various parameters such as accuracy, confusion matrix, precision, recall and F1 score.

General system architecture where model is replaced

The UCI-Epileptic Seizure Dataset is the set of data that is utilised for model performance. The original dataset consists of 100 files in 5 separate folders, each of which corresponds to a particular patient. 23.6-s recording of brain activity is stored in each file. A total of 4097 information/data points from the associated time series are sampled. Each data point represents the EEG recording’s value at a particular time point. There are 500 people in all, each having 4097 data points. Every 4097 information points, it is split and jumbled into 23 sets, each of which has 178 information points for 1 s and represents the value of the EEG recording at a particular time. As a result, there are now 23 (sets) × 500 (people) = 11,500 informational pieces, each of which comprises 178 information points for 1 s (a column), the last column represents the class as y {1,2,3,4,5}. The dataset therefore has 179 total columns, the first 178 of which are input vectors, and the 179th of which is a categorisation for patients (Table  1 ).

Dataset preprocessing

The dataset consists of 5 different classes mentioned as follows:

5—Keeping patient eyes open while taking a brain EEG graph

4—Keeping patient eyes closed while taking a brain EEG graph

3—EEG activity from a healthy part of the brain

2—EEG activity in the vicinity of the tumour

1—Seizure activity recording

All patients in classes 5, 4, 3 and 2 are those where there is no experience of seizures, according to an analysis of the dataset. The dataset is then encoded using the One-Hot method for binary classification, with all classes except 1 being transformed to 0 to indicate no seizures and 1 to indicate seizure sufferers.

ML and DL model selection

As the part of comparative study, there are a total of five models compared based on model evaluation parameters. Logistic regression, SVM classifier and KNN are traditional machine learning algorithm; as there is binary classification, these models were selected. The artificial neural network was used for complex analysis, and the last model was the LSTM-based neural network; with this model, other models were compared.

Results and discussion

Traditional machine learning model, model 1: logistic regression.

The regression algorithm for binary classification is used for detecting seizure activity. The model used test spilt as 33% of the dataset for validation. The training accuracy of the model is 66.92%, and the validation accuracy is 63.9%. Using validation data, as shown in Fig.  2 , the true-negative value percentage was 67.17%, true-positive 0.0%, false-positive rate 32.83% and false negative 0.0%. As the confusion matrix is shown in Fig.  2 , the classes are not perfectly classified, and it is not useful for this classification task.

Logistic regression confusion matrix

Model 2: SVM

For binary classification, SVM is used. The model is trained on 67% of the dataset. The validation accuracy is 97.2%, and the training accuracy is 98.09%. The model has a 0.0% true-positive classification on validation data. As shown in Fig.  3 , false positive 19.10%. The true negative was 80.90%, and the false negative was 0%. The validation data is 720 rows as positive class and 3025 as negative class. As the confusion matrix is shown in Fig.  3 , the classes are not perfectly classified, and it is not useful for this classification task.

SVM confusion matrix

Model 3: K-nearest neighbour

The KNN is used for the classification of seizure activity. The neighbour value is 5, and the distance formula is Euclidean distance. The training accuracy of the model is 93.61%, and the validation accuracy is 91.96% based on 33% data of the dataset. Using validation data, the true-negative percentage was 87.59% and the true positive as 0.0%. The false-negative value is 0.0%, and the false-positive value is 12.41% as shown in Fig.  4 . As the confusion matrix shown in Fig.  4 , the classes are not perfectly classified, and it is not useful for this classification task.

KNN confusion matrix

Deep neural network

Model 1: artificial neural network (ann).

The multiple-layer ANN model is implemented with four layers. The input shape for the model was 178.1 that is 178 data points. The first, second and third, layers with 32 neurons and Relu is the activation function, and the fourth layer with 2 neurons with softmax is the activation function. Adam optimiser was used as a model optimiser. Binary cross entropy is used as loss as there is binary classification. The total trainable parameters are 7906. The model is trained on 67% data. The model training accuracy was 98.9%, and the validation accuracy was 97% which is tested on 33% of testing dataset. It was observed a sudden increase in training accuracy while training the model after the first epoch from 0.91 to 0. 96% and a slow increase in accuracy shown in Fig.  5 .

ANN accuracy

The training loss is 0.03, and the validation loss is 0.08 as shown in Fig.  6 . Based on validation data, the confusion matrix is shown in Fig.  7 .

Confusion matrix

True negative (TN) is 78.89%, true positive (TP) is 18.16%, false positive is 0.81% and false negative is 2.13%. The model validation values for negative class (no seizure) precision were 0.97%, recall 0.99% and F1 score 0.98% and positive class (seizure activity) precision 0.96%, recall 0.89% and F1 score 0.92%. The total number of negative class data rows is 3025 and the positive class 770. As the system works for the medical domain, there is importance to the false-negative value which looks effective in ANN, so using the ANN model for this task is not that much good choice.

Proposed model

Model 1: proposed model—long short-term memory.

The multiple-layer LSTM model is implemented with three layers. The input shape for the model was 1178 that is 178 data points. The first LSTM layer with 64 neurons and Relu as activation function. The second LSTM layer with 32 neurons and Relu as an activation function. The last layer with 2 neurons with softmax as an activation function converts the output to a weighted sum to probability which sums to 1. Adam optimiser is used as an optimiser. Binary cross entropy is used as loss as there is binary classification. Total trainable parameters as 74,690. The model is trained on 67% data. The model training accuracy is 99.9%, and the validation accuracy is 97% which is tested on 33% of the dataset. It was observed a sudden increase in training accuracy while training the model after the first epoch from 0.91 to 0.96%, and a slow increase in accuracy is shown in Fig.  8 .

LSTM accuracy

The training loss is 0.006, and the validation loss is 0.106 as shown in Fig.  9 . Based on validation data, the confusion matrix is shown in Fig.  10 .

LSTM confusion matrix

True negative (TN) is 79.05%, true positive (TP) 18.23%, false-positive percentage 0.66% and false negative 2.06%. The model validation values for negative class (no seizure) precision 0.97%, recall 0.99%, and F1 score 0.98% and positive class (seizure activity) precision 0.97%, recall 0.90% and F1 score 0.93%. The total number of negative class data rows is 3025 and positive class 770.

Comparitive analysis

There are a total of five models trained and tested on the UCI-Epileptic Seizure Dataset. Each model was compared with other models based on several model evaluation parameters. In this study, precision, recall, training accuracy, validation accuracy, F1-Score and confusion matrix are the parameters considered.

The matrix below shows how much actual values as the same as predicted values by model, based on that true negative, false negatives, false positive and true positive are calculated.

Actual values

It tells how much the per cent model gives accurate values.

It reveals how many of the positive data points that the model identified as positive are truly positive.

The recall measures how accurately the model has identified real data items.

Table 2 gives a detailed comparison of all 5 models used in this study. All the models were trained and validated on the same dataset with validation data as 33% UCI dataset and 67% for training models. As shown in Table  2 , the positive class (seizure activity) have 770 rows for validation and 3025 as a negative class (no seizure). The logistic regression achieves training and validation accuracy of 66.92% and 63.9%, respectively, comparatively less than the SVM used on the same dataset. Similarly, with precision, the F1 score and recall value in the logistic regression algorithm show very less effect in seizure classification. Refer to the confusion matrix of SVM shown in Fig.  3 great training accuracy of 98.09% and validation accuracy of 97.23% but not able to predict classes as shown as true positive. The KNN showed training accuracy and validation accuracy of 93.61% and 91.96%, respectively, but not able to detect classes as the true positive rate is 0.0%.

The ANN model was able to classify true-positive rate (sensitivity) value with minimal false positive compared to SVM. ANN shown in Table  2 gives a precision value of 0.96, recall 0.89 and F1 score 0.92 for seizure signals and precision value 0.97, recall 0.99 and F1 score 0.98 for healthy signals. The proposed model LSTM was able to classify more accurately than ANN with very minimal difference in training and validation accuracy. The proposed model was able to achieve 99.88% accuracy on training data and a validation accuracy of 97.1% as compared to ANN validation accuracy of 97.0%. LSTM shown in Table  2 gives a precision value of 0.96, recall 0.90 and F1 score 0.93 for seizure signals and precision value 0.97, recall 0.99 and F1 score 0.98 for healthy signals. Based on the overall comparison, LSTM-based model performs better.

Conclusions

As the proposed system intended to classify the healthy person’s brain EEG signal and seizure patient brain EEG signal, the system classifies the signal data with the LSTM model with a validation accuracy of 97% and false negative 2.06%. As shown in Table  2 , the conventional machine learning algorithms logistic regression, SVM and KNN achieve good accuracy but not work fine in classification. Also, considering the ANN achieves good model evaluation parameters but somewhat less precise in false negative area. In conclusion, the proposed model works better as compared to other models used in this study. The system can be useful in epileptic seizure detection.

The proposed system currently works good in binary classification. There are also types of epileptic seizure that can be detected precisely as it deals with the medical domain. The dataset used in this study is somewhat insufficient to train model, and also dataset is unbalanced as the other categories are converted using one-hot encoding for binary classification. Also, in the proposed model, there is a scope of improvement in the false-negative section. The limitation of the proposed system is that it needs to test for multiclass classification of epilepsy seizures. The similar work has been carried out in [ 24 ]. Also, work needs to be tested with other datasets like state-of-the-art work [ 25 , 26 , 27 ]. The other alternatives to test the results are to use EEG datasets like CHB-MIT, TUH EEG Corpus and Bonn University with methods like CNN, SofMax and Bi-LSTM to improve false negatives.

Availability of data and materials

Data is publicly available on Kaggle and UCI Repository.

Abbreviations

Machine learning

Deep learning

University of California Irvine

Long short-term memory

World Health Organization

• Electroencephalogram

Empirical mode decomposition

Convolutional neural network

k -Nearest neighbours

Random forest

Support vector machine

Artificial neural networks

Decision tree

Principal component analysis

Least squares support vector machine

Kunekar PR, Gupta M, Agarwal B (2020) Deep learning with multi modal ensemble fusion for epilepsy diagnosis. 2020 3rd International Conference on Emerging Technologies in Computer Engineering: Machine Learning and Internet of Things (ICETCE). pp 80–84. https://doi.org/10.1109/ICETCE48199.2020.9091742

Chapter   Google Scholar  

Kunekar PR, Gupta M, Agarwal B (2019) Detection and analysis of life style based diseases in early phase of life: a survey. In: Somani A, Ramakrishna S, Chaudhary A, Choudhary C, Agarwal B (eds) Emerging technologies in computer engineering: microservices in big data analytics. ICETCE 2019. Communications in Computer and Information Science, vol 985. Springer, Singapore. https://doi.org/10.1007/978-981-13-8300-7_6

Kumar Boddu RS, Chakravarthi DS, Venkateswararao N, Chakravarthy DSK, Devarajan A, Kunekar PR (2021) The effects of artificial intelligence and medical technology on the life of human. J Pharm Res Int 33(50A):25–32. Article no.JPRI.76516, ISSN: 2456-9119

Article   Google Scholar  

Arora B, Jadhav P, Sulaiman RB, Kareem A, Kunekar P, Pant B (2022) Integrating artificial intelligence and deep learning for enhanced medical innovation. 2022 5th International Conference on Contemporary Computing and Informatics (IC3I). pp 327–331. https://doi.org/10.1109/IC3I56241.2022.10073054

Kunekar P, Vaishnav R, Kalaivani E, Gangodkar D, Kaur C, Dhanraj JA (2022) Applications of machine learning techniques in detecting skin cancer. 2022 5th International Conference on Contemporary Computing and Informatics (IC3I). pp 265–270. https://doi.org/10.1109/IC3I56241.2022.10072834

Chakrapani IS, Tyagi N, Tyagi S, Kunekar P, Padmaja DL, Pant K (2022) Applications of Deep Learning (DL) techniques in detecting breast cancer and malignant cells. 2022 5th International Conference on Contemporary Computing and Informatics (IC3I). pp 217–222. https://doi.org/10.1109/IC3I56241.2022.10072720

Usman SM, Usman M, Fong S (2017) Epileptic seizures prediction using machine learning methods. Comput Math Methods Med 2017:9074759

Article   MathSciNet   Google Scholar  

Zhou M, Tian C, Cao R, Wang B, Niu Y, Hu T, Guo H, Xiang J (2018) Epileptic seizure detection based on EEG signals and CNN. Front Neuroinform 12:95

Serna J, Paternina MRA, Zamora-Méndez ARK. Tripathy R, Pachori R (2020) EEG-Rhythm Specific Taylor-Fourier Filter Bank Implemented With O-Splines for the Detection of Epilepsy Using EEG Signals. IEEE Sens J. https://doi.org/10.1109/JSEN.2020.2976519

Raghu S, Sriraam N, Temel Y, Rao SV, Kubben PL (2020) EEG based multi-class seizure type classification using convolutional neural network and transfer learning. Neural Netw 124:202–212

Ludwig SA (2020) Multi-label classification for epileptic seizure recognition: deep neural network ensemble versus Choquet fuzzy integral fusion. 2020 IEEE Symposium Series on Computational Intelligence (SSCI). pp 836–841

Singh N, Dehuri S (2020) Multiclass classification of EEG signal for epilepsy detection using DWT based SVD and fuzzy kNN classifier. Intell Decis Technol. 14:1–14. https://doi.org/10.3233/IDT-190043

Aayesha Z, Qureshi M, Afzaal M, Qureshi M, Fayaz M (2021) Machine learning-based EEG signals classification model for epileptic seizure detection. Multimed Tools Appl. 80:1–29. https://doi.org/10.1007/s11042-021-10597-6

Abdelhameed A, Bayoumi M (2021) A deep learning approach for automatic seizure detection in children with epilepsy. Front Comput Neurosci 15:650050. United States

Nahzat S, Yaganoglu M (2021) Classification of epileptic seizure dataset using different machine learning algorithms and PCA feature reduction technique. J Invest Eng Technol 4(2):47–60

Google Scholar  

Shoeibi A, Khodatars M, Ghassemi N, Jafari M, Moridian P, Alizadehsani R et al (2021) Epileptic Seizures Detection Using Deep Learning Techniques: A Review. Int J Environ Res Public Health. 18(11):5780. https://doi.org/10.3390/ijerph18115780

Raibag MAA, Franklin JV, Sarkar R (2022) Multi-feature learning model for epilepsy classification supervised by a highly robust heterogeneous deep ensemble. Turk J Comput Math Educ 13:273–284

Torse DA, Khanai R (2022) Classification of epileptic seizures using ensemble empirical mode decomposition and least squares support vector machine. 2021 International Conference on Computer Communication and Informatics (ICCCI). pp 1–5

Natu M, Bachute M, Gite S, Kotecha K, Vidyarthi A (2022) Review on epileptic seizure prediction: machine learning and deep learning approaches. Comput Math Methods Med 2022:7751263

George F et al (2022) Epileptic seizure prediction using EEG images. 2020 International Conference on Communication and Signal Processing (ICCSP). pp 1595–1598

Ouichka O, Echtioui A, Hamam H (2022) Deep Learning Models for Predicting Epileptic Seizures Using iEEG Signals. Electronics. 11(4):605. https://doi.org/10.3390/electronics11040605

Chirasani SKR, Manikandan S (2022) A deep neural network for the classification of epileptic seizures using hierarchical attention mechanism. Appl Soft Comput 26:5389–5397

Jiang Y, Yao Lu, Yang L (2022) An epileptic seizure prediction model based on a time-wise attention simulation module and a pretrained ResNet. Methods 202:117–126

Oliva JT, Rosa JLG (2021) Binary and multiclass classifiers based on multitaper spectral features for epilepsy detection. Biomed Signal Process Control 66:102469

Raghu S, Sriraam N, Temel Y, Rao SV, Kubben PL (2020) EEG based multi-class seizure type classification using convolutional neural network and transfer learning. Neural Netw. 124:202–212. https://doi.org/10.1016/j.neunet.2020.01.017 . Epub 2020 Jan 25

Ilias L, Askounis D, Psarras J (2022) Multimodal detection of epilepsy with deep neural networks. Expert Syst Appl. 213:119010. https://doi.org/10.1016/j.eswa.2022.119010

Yao X, Cheng Q, Zhang GQ (2019) A novel independent RNN approach to classification of seizures against non-seizures. https://arxiv.org/abs/1903.09326

Download references

Acknowledgements

Not applicable.

Author information

Authors and affiliations.

Department of CSE, Swami Keshvanand Institute of Technology, Management & Gramothan, Jaipur, Rajasthan, India

Pankaj Kunekar & Mukesh Kumar Gupta

BITS Pilani, Dubai Campus, Dubai, UAE

Pramod Gaur

You can also search for this author in PubMed   Google Scholar

Contributions

The original idea of the research work is of PK. Literature work and implementation work are done by PK. Reviews, suggestions and inputs to research work are done by MG and PG. Also, MG and PG have contributed to the implementation work. Results are validated and verified by PK, MG and PG. All authors have read and approved the manuscript.

Corresponding author

Correspondence to Pankaj Kunekar .

Ethics declarations

Competing interests.

The authors declare that they have no competing interests.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Kunekar, P., Gupta, M.K. & Gaur, P. Detection of epileptic seizure in EEG signals using machine learning and deep learning techniques. J. Eng. Appl. Sci. 71 , 21 (2024). https://doi.org/10.1186/s44147-023-00353-y

Download citation

Received : 09 May 2023

Accepted : 23 December 2023

Published : 19 January 2024

DOI : https://doi.org/10.1186/s44147-023-00353-y

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Epilepsy analysis
• Epileptic seizure detection
• Comparative analysis
• UCI dataset

Research progress of epileptic seizure prediction methods based on EEG

• Review Paper
• Published: 07 May 2024

Cite this article

• Zhongpeng Wang 1 , 2 ,
• Xiaoxin Song   ORCID: orcid.org/0009-0008-0176-2310 1 ,
• Long Chen 1 , 2 ,
• Jinxiang Nan 1 ,
• Yulin Sun 1 ,
• Meijun Pang 1 , 2 ,
• Kuo Zhang 1 , 2 ,
• Xiuyun Liu 1 , 2 &
• Dong Ming 1 , 2  

Explore all metrics

At present, at least 30% of refractory epilepsy patients in the world cannot be effectively controlled and treated. The suddenness and unpredictability of seizures greatly affect the physical and mental health and even the life safety of patients, and the realization of early prediction of seizures and the adoption of interventions are of great significance to the improvement of patients’ quality of life. In this paper, we firstly introduce the design process of EEG-based seizure prediction methods, introduce several databases commonly used in the research, and summarize the commonly used methods in pre-processing, feature extraction, classification and identification, and post-processing. Then, based on scalp EEG and intracranial EEG respectively, we reviewed the current status of epileptic seizure prediction research from five commonly used feature analysis methods, and make a comprehensive evaluation of both. Finally, this paper describes the reasons why the current algorithms cannot be applied to the clinic, summarizes their limitations, and gives corresponding suggestions, aiming to provide improvement directions for subsequent research. In addition, deep learning algorithms have emerged in recent years, and this paper also compares the advantages and disadvantages of deep learning algorithms with traditional machine learning methods, in the hope of providing researchers with new technologies and new ideas and making significant breakthroughs in the field of epileptic seizure prediction.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price includes VAT (Russian Federation)

Instant access to the full article PDF.

Rent this article via DeepDyve

Institutional subscriptions

Similar content being viewed by others

Deep learning techniques for classification of electroencephalogram (EEG) motor imagery (MI) signals: a review

EEG seizure detection: concepts, techniques, challenges, and future trends

Emerging Trends in EEG Signal Processing: A Systematic Review

Aarabi A, Fazel-Rezai R, Aghakhani Y (2009) EEG seizure prediction: measures and challenges. In: 2009 Annual international conference of the IEEE engineering in medicine and biology society, Minneapolis, MN, USA, 03–06 September 2009

Aarabi A, He B (2017) Seizure prediction in patients with focal hippocampal epilepsy. Clin Neurophysiol 128(7):1299–1307. https://doi.org/10.1016/j.clinph.2017.04.026

Article   PubMed   PubMed Central   Google Scholar  

Abbasi B, Goldenholz DM (2019) Machine learning applications in epilepsy. Epilepsia 60(10):2037–2047. https://doi.org/10.1111/epi.16333

Acharya UR, Vinitha Sree S, Swapna G, Martis RJ, Suri JS (2013) Automated EEG analysis of epilepsy: a review. Knowl-Based Syst 45:147–165. https://doi.org/10.1016/j.knosys.2013.02.014

Article   Google Scholar  

Acharya UR, Hagiwara Y, Adeli H (2018a) Automated seizure prediction. Epilepsy Behav 88:251–261. https://doi.org/10.1016/j.yebeh.2018.09.030

Article   PubMed   Google Scholar  

Acharya UR, Oh SL, Hagiwara Y, Tan JH, Adeli H (2018b) Deep convolutional neural network for the automated detection and diagnosis of seizure using EEG signals. Comput Biol Med 100:270–278. https://doi.org/10.1016/j.compbiomed.2017.09.017

Alexandre Teixeira C, Direito B, Bandarabadi M, Le Van QM, Valderrama M, Schelter B, Schulze-Bonhage A, Navarro V, Sales F, Dourado A (2014) Epileptic seizure predictors based on computational intelligence techniques: a comparative study with 278 patients. Comput Methods Programs Biomed 114(3):324–336. https://doi.org/10.1016/j.cmpb.2014.02.007

Alotaiby TN, Alshebeili SA, Alotaibi FM, Alrshoud SR (2017) Epileptic seizure prediction using CSP and LDA for scalp EEG signals. Comput Intell Neurosci 2017:1240323. https://doi.org/10.1155/2017/1240323

Bandarabadi M, Teixeira CA, Rasekhi J, Dourado A (2015) Epileptic seizure prediction using relative spectral power features. Clin Neurophysiol 126(2):237–248. https://doi.org/10.1016/j.clinph.2014.05.022

Beniczky S, Karoly P, Nurse E, Ryvlin P, Cook M (2021) Machine learning and wearable devices of the future. Epilepsia 62(Suppl 2):S116–S124. https://doi.org/10.1111/epi.16555

Bishop CM (2006) Pattern Recognition and Machine Learning. Springer New York, New York. https://doi.org/10.1007/978-0-387-45528-0

Book   Google Scholar  

Blanco S, Garay A, Coulombie D (2013) Comparison of frequency bands using spectral entropy for epileptic seizure prediction. ISRN Neurol 2013:1–5. https://doi.org/10.1155/2013/287327

Bonn E a U. Epileptologie Bonn / Forschung / AG Lehnertz / EEG Data Download n.d. https://www.ukbonn.de/epileptologie/?idcat=193&lang=3

Brázdil M, Pail M, Halámek J, Plešinger F, Cimbálník J, Roman R, Klimeš P, Daniel P, Chrastina J, Brichtová E, Rektor I, Worrell GA, Jurák P (2017) Very high-frequency oscillations: Novel biomarkers of the epileptogenic zone. Ann Neurol 82(2):299–310. https://doi.org/10.1002/ana.25006

Burges CJC (1998) A tutorial on support vector machines for pattern recognition. Data Min Knowl Disc 2(2):121–167. https://doi.org/10.1023/a:1009715923555

Chen H, Ji T, Zhan X, Liu X, Yu G, Wang W, Jiang Y, Zhou X-H, Ullah I (2022) An explainable statistical method for seizure prediction using brain functional connectivity from EEG. Comput Intell Neurosci 2022:1–8. https://doi.org/10.1155/2022/2183562

Chu H, Chung CK, Jeong W, Cho KH (2017) Predicting epileptic seizures from scalp EEG based on attractor state analysis. Comput Methods Programs Biomed 143:75–87. https://doi.org/10.1016/j.cmpb.2017.03.002

Competition RP (2014) American Epilepsy Society seizure prediction challenge. https://www.kaggle.com/competitions/seizure-prediction/data

Cui S, Duan L, Qiao Y, Xiao Y (2018) Learning EEG synchronization patterns for epileptic seizure prediction using bag-of-wave features. J Ambient Intell Humaniz Comput (prepublish). https://doi.org/10.1007/s12652-018-1000-3

D’Alessandro M, Esteller R, Vachtsevanos G, Hinson A, Echauz J, Litt B (2003) Epileptic seizure prediction using hybrid feature selection over multiple intracranial EEG electrode contacts: a report of four patients. IEEE Trans Biomed Eng 50(5):603–615. https://doi.org/10.1109/tbme.2003.810706

Delamont RS, Walker MC (2011) Pre-ictal autonomic changes. Epilepsy Res 97(3):267–272. https://doi.org/10.1016/j.eplepsyres.2011.10.016

Detti P, Vatti G, Zabalo Manrique de Lara G (2020) EEG synchronization analysis for seizure prediction: a study on data of noninvasive recordings. Processes 8(7). https://doi.org/10.3390/pr8070846

Direito B, Teixeira CA, Sales F, Castelo-Branco M, Dourado A (2017) A realistic seizure prediction study based on multiclass SVM. Int J Neural Syst 27(3):1750006. https://doi.org/10.1142/s012906571750006x

Duun-Henriksen J, Kjaer TW, Madsen RE, Remvig LS, Thomsen CE, Sorensen HB (2012) Channel selection for automatic seizure detection. Clin Neurophysiol 123(1):84–92. https://doi.org/10.1016/j.clinph.2011.06.001

Elger CE, Lehnertz K (1998) Seizure prediction by non-linear time series analysis of brain electrical activity. Eur J Neurosci 10(2):786–789. https://doi.org/10.1046/j.1460-9568.1998.00090.x

Article   CAS   PubMed   Google Scholar  

EPMoghaddam D, Sheth SA, Haneef Z, Gavvala J, Aazhang B (2022) Epileptic seizure prediction using spectral width of the covariance matrix. J Neural Eng 19(2). https://doi.org/10.1088/1741-2552/ac6063

Fei K, Wang W, Yang Q, Tang S (2017) Chaos feature study in fractional Fourier domain for preictal prediction of epileptic seizure. Neurocomputing 249:290–298. https://doi.org/10.1016/j.neucom.2017.04.019

Freiburg UO (2003) Seizure prediction project Freiburg. https://epilepsy.uni-freiburg.de/

Freiman TM, Eismann-Schweimler J, Frotscher M (2011) Granule cell dispersion in temporal lobe epilepsy is associated with changes in dendritic orientation and spine distribution. Exp Neurol 229(2):332–338. https://doi.org/10.1016/j.expneurol.2011.02.017

Gao Y, Liu A, Wang L, Qian R, Chen X (2023) A self-interpretable deep learning model for seizure prediction using a multi-scale prototypical part network. IEEE Trans Neural Syst Rehabil Eng 31:1847–1856. https://doi.org/10.1109/tnsre.2023.3260845

Ghosh A, Sarkar A, Das T, Basak P (2017) Pre-ictal epileptic seizure prediction based on ECG signal analysis. In: 2017 2nd International conference for convergence in technology (I2CT), Mumbai, India, 07–09 April 2017

Goldberger AL, Amaral LA, Glass L, Hausdorff JM, Ivanov PC, Mark RG, Mietus JE, Moody GB, Peng CK, Stanley HE (2000) PhysioBank, PhysioToolkit, and PhysioNet: components of a new research resource for complex physiologic signals. Circulation 101(23):E215–E220. https://doi.org/10.1161/01.cir.101.23.e215

Han C, Peng F, Chen C, Li W, Zhang X, Wang X, Zhou W (2021) Research progress of epileptic seizure predictions based on electroencephalogram signals. Sheng Wu Yi Xue Gong Cheng Xue Za Zhi 38(6):1193–1202. https://doi.org/10.7507/1001-5515.202105052

Hao J, Cui Y, Niu B, Yu L, Lin Y, Xia Y, Yao D, Guo D (2021) Roles of very fast ripple (500–1000[Formula: see text]Hz) in the hippocampal network during status epilepticus. Int J Neural Syst 31(4):2150002. https://doi.org/10.1142/s0129065721500027

Hassan AR, Subasi A (2016) Automatic identification of epileptic seizures from EEG signals using linear programming boosting. Comput Methods Programs Biomed 136:65–77. https://doi.org/10.1016/j.cmpb.2016.08.013

Honghong H, Feng Z, Liangfu L, Xiaopeng S (2023) A review of the application of neural networks in epileptic seizure prediction. J Front Comput Sci Technol 17(11):2543–2556. https://doi.org/10.3778/j.issn.1673-9418.2302001

Howbert JJ, Patterson EE, Stead SM, Brinkmann B, Vasoli V, Crepeau D, Vite CH, Sturges B, Ruedebusch V, Mavoori J, Leyde K, Sheffield WD, Litt B, Worrell GA (2014) Forecasting seizures in dogs with naturally occurring epilepsy. PLoS ONE 9(1):e81920. https://doi.org/10.1371/journal.pone.0081920

Article   CAS   PubMed   PubMed Central   Google Scholar  

Iasemidis LD, Sackellares JC, Zaveri HP, Williams WJ (1990) Phase space topography and the Lyapunov exponent of electrocorticograms in partial seizures. Brain Topogr 2(3):187–201. https://doi.org/10.1007/bf01140588

Iasemidis LD, Deng-Shan S, Chaovalitwongse W, Sackellares JC, Pardalos PM, Principe JC, Carney PR, Prasad A, Veeramani B, Tsakalis K (2003) Adaptive epileptic seizure prediction system. IEEE Trans Biomed Eng 50(5):616–627. https://doi.org/10.1109/tbme.2003.810689

Ibrahim F, Abd-Elateif El-Gindy S, El-Dolil SM, El-Fishawy AS, El-Rabaie E-SM, Dessouky MI, Eldokany IM, Alotaiby TN, Alshebeili SA, Abd El-Samie FE (2019) A statistical framework for EEG channel selection and seizure prediction on mobile. Int J Speech Technol 22(1):191–203. https://doi.org/10.1007/s10772-018-09565-7

Islam MS, El-Hajj AM, Alawieh H, Dawy Z, Abbas N, El-Imad J (2020) EEG mobility artifact removal for ambulatory epileptic seizure prediction applications. Biomed Signal Process Control 55. https://doi.org/10.1016/j.bspc.2019.101638

Jacobs J, LeVan P, Chander R, Hall J, Dubeau F, Gotman J (2008) Interictal high-frequency oscillations (80–500 Hz) are an indicator of seizure onset areas independent of spikes in the human epileptic brain. Epilepsia 49(11):1893–1907. https://doi.org/10.1111/j.1528-1167.2008.01656.x

Jacobs J, Zijlmans M, Zelmann R, Chatillon CE, Hall J, Olivier A, Dubeau F, Gotman J (2010) High-frequency electroencephalographic oscillations correlate with outcome of epilepsy surgery. Ann Neurol 67(2):209–220. https://doi.org/10.1002/ana.21847

Jemal I, Mitiche A, Mezghani N (2021) A study of EEG feature complexity in epileptic seizure prediction. Appl Sci 11(4). https://doi.org/10.3390/app11041579

Khan H, Marcuse L, Fields M, Swann K, Yener B (2018) Focal onset seizure prediction using convolutional networks. IEEE Trans Biomed Eng 65(9):2109–2118. https://doi.org/10.1109/tbme.2017.2785401

Klatt J, Feldwisch-Drentrup H, Ihle M, Navarro V, Neufang M, Teixeira C, Adam C, Valderrama M, Alvarado-Rojas C, Witon A, Le Van QM, Sales F, Dourado A, Timmer J, Schulze-Bonhage A, Schelter B (2012) The EPILEPSIAE database: an extensive electroencephalography database of epilepsy patients. Epilepsia 53(9):1669–1676. https://doi.org/10.1111/j.1528-1167.2012.03564.x

Le Douget JE, Fouad A, Maskani Filali M, Pyrzowski J, Le Van QM (2017) Surface and intracranial EEG spike detection based on discrete wavelet decomposition and random forest classification. Annu Int Conf IEEE Eng Med Biol Soc 2017:475–478. https://doi.org/10.1109/embc.2017.8036865

Lekshmy HO, Panickar D, Harikumar S (2022) Comparative analysis of multiple machine learning algorithms for epileptic seizure prediction. J Phys Conf Ser 2161(1). https://doi.org/10.1088/1742-6596/2161/1/012055

Li S, Zhou W, Yuan Q, Liu Y (2013) Seizure prediction using spike rate of intracranial EEG. IEEE Trans Neural Syst Rehabil Eng 21(6):880–886. https://doi.org/10.1109/tnsre.2013.2282153

Li Y, Liu Y, Guo YZ, Liao XF, Hu B, Yu T (2022) Spatio-temporal-spectral hierarchical graph convolutional network with semisupervised active learning for patient-specific seizure prediction. IEEE Trans Cybern 52(11):12189–12204. https://doi.org/10.1109/tcyb.2021.3071860

Litt B, Echauz J (2002) Prediction of epileptic seizures. Lancet Neurol 1(1):22–30. https://doi.org/10.1016/s1474-4422(02)00003-0

Lu L, Zhang F, Wu Y, Ma S, Zhang X, Ni G (2022) A multi-frame network model for predicting seizure based on sEEG and iEEG data. Front Comput Neurosci 16:1059565. https://doi.org/10.3389/fncom.2022.1059565

Maimaiti B, Meng H, Lv Y, Qiu J, Zhu Z, Xie Y, Li Y, Yu C, Zhao W, Liu J, Li M (2022) An overview of EEG-based machine learning methods in seizure prediction and opportunities for neurologists in this field. Neuroscience 481:197–218. https://doi.org/10.1016/j.neuroscience.2021.11.017

Mirowski P, Madhavan D, LeCun Y, Kuzniecky R (2009) Classification of patterns of EEG synchronization for seizure prediction. Clin Neurophysiol 120(11):1927–1940. https://doi.org/10.1016/j.clinph.2009.09.002

Mormann F, Lehnertz K, David P, Elger CE (2000) Mean phase coherence as a measure for phase synchronization and its application to the EEG of epilepsy patients. Physica D 144(3–4):358–369. https://doi.org/10.1016/s0167-2789(00)00087-7

Mormann F, Kreuz T, Rieke C, Andrzejak RG, Kraskov A, David P, Elger CE, Lehnertz K (2005) On the predictability of epileptic seizures. Clin Neurophysiol 116(3):569–587. https://doi.org/10.1016/j.clinph.2004.08.025

Nabil D, Benali R, Reguig FB (2020) Epileptic seizure recognition using EEG wavelet decomposition based on nonlinear and statistical features with support vector machine classification. Biomed Eng Biomed Tech 65(2):133–148. https://doi.org/10.1515/bmt-2018-0246

Netoff T, Park Y, Parhi K (2009) Seizure prediction using cost-sensitive support vector machine. Annu Int Conf IEEE Eng Med Biol Soc 2009:3322–3325. https://doi.org/10.1109/iembs.2009.5333711

Ngamga EJ, Bialonski S, Marwan N, Kurths J, Geier C, Lehnertz K (2016) Evaluation of selected recurrence measures in discriminating pre-ictal and inter-ictal periods from epileptic EEG data. Phys Lett A 380(16):1419–1425. https://doi.org/10.1016/j.physleta.2016.02.024

Article   CAS   Google Scholar  

Osmond E, Billetop A, Jary S, Likeman M, Thoresen M, Luyt K (2014) Neonatal seizures: magnetic resonance imaging adds value in the diagnosis and prediction of neurodisability. Acta Paediatr 103(8):820–826. https://doi.org/10.1111/apa.12583

Oyrer J, Maljevic S, Scheffer IE, Berkovic SF, Petrou S, Reid CA (2018) Ion channels in genetic epilepsy: from genes and mechanisms to disease-targeted therapies. Pharmacol Rev 70(1):142–173. https://doi.org/10.1124/pr.117.014456

Parhi KK, Zhang Z (2019) Discriminative ratio of spectral power and relative power features derived via frequency-domain model ratio with application to seizure prediction. IEEE Trans Biomed Circuits Syst 13(4):645–657. https://doi.org/10.1109/tbcas.2019.2917184

Park Y, Luo L, Parhi KK, Netoff T (2011) Seizure prediction with spectral power of EEG using cost-sensitive support vector machines. Epilepsia 52(10):1761–1770. https://doi.org/10.1111/j.1528-1167.2011.03138.x

Patel AD, Moss R, Rust SW, Patterson J, Strouse R, Gedela S, Haines J, Lin SM (2016) Patient-centered design criteria for wearable seizure detection devices. Epilepsy Behav 64(Pt A):116–121. https://doi.org/10.1016/j.yebeh.2016.09.012

Peters B (2022) Risks and complications of seizures. https://www.verywellhealth.com/risks-and-complications-of-seizures-4685790

Qin Y, Zheng H, Chen W, Qin Q, Han C, Che Y (2020). Patient-specific seizure prediction with scalp EEG using convolutional neural network and extreme learning machine. In: 2020 39th Chinese control conference (CCC), Shenyang, China, 27–29 July 2020

Rajkomar A, Dean J, Kohane I (2019) Machine learning in medicine. N Engl J Med 380(14):1347–1358. https://doi.org/10.1056/NEJMra1814259

Rasekhi J, Mollaei MR, Bandarabadi M, Teixeira CA, Dourado A (2013) Preprocessing effects of 22 linear univariate features on the performance of seizure prediction methods. J Neurosci Methods 217(1–2):9–16. https://doi.org/10.1016/j.jneumeth.2013.03.019

Savadkoohi M, Oladunni T, Thompson L (2020) A machine learning approach to epileptic seizure prediction using Electroencephalogram (EEG) Signal. Biocybern Biomed Eng 40(3):1328–1341. https://doi.org/10.1016/j.bbe.2020.07.004

Schelter B, Winterhalder M, Maiwald T, Brandt A, Schad A, Schulze-Bonhage A, Timmer J (2006) Testing statistical significance of multivariate time series analysis techniques for epileptic seizure prediction. Chaos 16(1):013108. https://doi.org/10.1063/1.2137623

Schulze-Bonhage A, Sales F, Wagner K, Teotonio R, Carius A, Schelle A, Ihle M (2010) Views of patients with epilepsy on seizure prediction devices. Epilepsy Behav 18(4):388–396. https://doi.org/10.1016/j.yebeh.2010.05.008

Shahidi Zandi A, Tafreshi R, Javidan M, Dumont GA (2013) Predicting epileptic seizures in scalp EEG based on a variational Bayesian Gaussian mixture model of zero-crossing intervals. IEEE Trans Biomed Eng 60(5):1401–1413. https://doi.org/10.1109/tbme.2012.2237399

Sharif B, Jafari AH (2017) Prediction of epileptic seizures from EEG using analysis of ictal rules on Poincaré plane. Comput Methods Programs Biomed 145:11–22. https://doi.org/10.1016/j.cmpb.2017.04.001

Slimen IB, Boubchir L, Seddik H (2020) Epileptic seizure prediction based on EEG spikes detection of ictal-preictal states. J Biomed Res 34(3). https://doi.org/10.7555/jbr.34.20190097

Sun Y, Jin W, Si X, Zhang X, Cao J, Wang L, Yin S, Ming D (2022) Continuous seizure detection based on transformer and long-term iEEG. IEEE J Biomed Health Inform 26(11):5418–5427. https://doi.org/10.1109/jbhi.2022.3199206

Sun B, Lv J-J, Rui L-G, Yang Y-X, Chen Y-G, Ma C, Gao Z-K (2021) Seizure prediction in scalp EEG based channel attention dual-input convolutional neural network. Physica A Stat Mech Appl 584. https://doi.org/10.1016/j.physa.2021.126376

Tamanna T, Rahman M A, Sultana S, Haque M H, Parvez MZ (2021) Predicting seizure onset based on time-frequency analysis of EEG signals. Chaos Solitons Fractals 145. https://doi.org/10.1016/j.chaos.2021.110796

Tomlinson SB, Porter BE, Marsh ED (2017) Interictal network synchrony and local heterogeneity predict epilepsy surgery outcome among pediatric patients. Epilepsia 58(3):402–411. https://doi.org/10.1111/epi.13657

Truong ND, Nguyen AD, Kuhlmann L, Bonyadi MR, Yang J, Ippolito S, Kavehei O (2018) Convolutional neural networks for seizure prediction using intracranial and scalp electroencephalogram. Neural Netw 105:104–111. https://doi.org/10.1016/j.neunet.2018.04.018

Tsiouris Κ, Pezoulas VC, Zervakis M, Konitsiotis S, Koutsouris DD, Fotiadis DI (2018) A long short-term memory deep learning network for the prediction of epileptic seizures using EEG signals. Comput Biol Med 99:24–37. https://doi.org/10.1016/j.compbiomed.2018.05.019

Usman SM, Khalid S, Akhtar R, Bortolotto Z, Bashir Z, Qiu H (2019a) Using scalp EEG and intracranial EEG signals for predicting epileptic seizures: review of available methodologies. Seizure 71:258–269. https://doi.org/10.1016/j.seizure.2019.08.006

Usman SM, Khalid S, Akhtar R, Bortolotto Z, Bashir Z, Qiu HY (2019b) Using scalp EEG and intracranial EEG signals for predicting epileptic seizures: review of available methodologies. Seizure-Eur J Epilepsy 71:258–269. https://doi.org/10.1016/j.seizure.2019.08.006

Usman SM, Latif S, Beg A (2019c) Principle components analysis for seizures prediction using wavelet transform. Int J Adv Appl Sci 6(3):50–55. https://doi.org/10.21833/ijaas.2019.03.008

Varotto G, Tassi L, Franceschetti S, Spreafico R, Panzica F (2012) Epileptogenic networks of type II focal cortical dysplasia: a stereo-EEG study. Neuroimage 61(3):591–598. https://doi.org/10.1016/j.neuroimage.2012.03.090

Velmurugan J, Nagarajan SS, Mariyappa N, Mundlamuri RC, Raghavendra K, Bharath RD, Saini J, Arivazhagan A, Rajeswaran J, Mahadevan A, Malla BR, Satishchandra P, Sinha S (2019) Magnetoencephalography imaging of high frequency oscillations strengthens presurgical localization and outcome prediction. Brain 142(11):3514–3529. https://doi.org/10.1093/brain/awz284

Vezzani A, French J, Bartfai T, Baram TZ (2011a) The role of inflammation in epilepsy. Nat Rev Neurol 7(1):31–40. https://doi.org/10.1038/nrneurol.2010.178

Vezzani A, Maroso M, Balosso S, Sanchez MA, Bartfai T (2011b) IL-1 receptor/Toll-like receptor signaling in infection, inflammation, stress and neurodegeneration couples hyperexcitability and seizures. Brain Behav Immun 25(7):1281–1289. https://doi.org/10.1016/j.bbi.2011.03.018

Wang G, Wang D, Du C, Li K, Zhang J, Liu Z, Tao Y, Wang M, Cao Z, Yan X (2020a) Seizure Prediction Using Directed Transfer Function and Convolution Neural Network on Intracranial EEG. IEEE Trans Neural Syst Rehabil Eng 28(12):2711–2720. https://doi.org/10.1109/tnsre.2020.3035836

Wang Y, Shi Y, Cheng Y, He Z, Wei X, Chen Z, Zhou Y (2023) A spatiotemporal graph attention network based on synchronization for epileptic seizure prediction. IEEE J Biomed Health Inform 27(2):900–911. https://doi.org/10.1109/jbhi.2022.3221211

Wang Y, Zhou D, Yang X, Xu X, Ren L, Yu T, Zhou W, Shao X, Yang Z, Wang S, Cao D, Liu C, Kwan SY, Xiang J (2020b) Expert consensus on clinical applications of high-frequency oscillations in epilepsy. Acta Epileptol 2(1). https://doi.org/10.1186/s42494-020-00018-w

WHO (2023) World Health Organization. Epilepsy. https://www.who.int/news-room/fact-sheets/detail/epilepsy

Williamson JR, Bliss DW, Browne DW, Narayanan JT (2012) Seizure prediction using EEG spatiotemporal correlation structure. Epilepsy Behav 25(2):230–238. https://doi.org/10.1016/j.yebeh.2012.07.007

Wong M, Guo D (2013) Dendritic spine pathology in epilepsy: cause or consequence? Neuroscience 251:141–150. https://doi.org/10.1016/j.neuroscience.2012.03.048

Wu G, Li Z, Zhang Y, Dong X, Ye L (2019) Study of feature extraction algorithms for epileptic seizure prediction based on SVM. In: Communications, signal processing, and systems pp 2370–2377. https://doi.org/10.1007/978-981-10-6571-2_289

Wu J, Zhou T, Li T (2020) Detecting epileptic seizures in EEG signals with complementary ensemble empirical mode decomposition and extreme gradient boosting. Entropy (Basel), 22(2). https://doi.org/10.3390/e22020140

Xiang J, Li C, Li H, Cao R, Wang B, Han X, Chen J (2015) The detection of epileptic seizure signals based on fuzzy entropy. J Neurosci Methods 243:18–25. https://doi.org/10.1016/j.jneumeth.2015.01.015

Yang S, Li B, Zhang Y, Duan M, Liu S, Zhang Y, Feng X, Tan R, Huang L, Zhou F (2020) Selection of features for patient-independent detection of seizure events using scalp EEG signals. Comput Biol Med 119:103671. https://doi.org/10.1016/j.compbiomed.2020.103671

Yang X, Zhao J, Sun Q, Lu J, Ma X (2021) An effective dual self-attention residual network for seizure prediction. IEEE Trans Neural Syst Rehabil Eng 29:1604–1613. https://doi.org/10.1109/tnsre.2021.3103210

Yu Z, Nie W, Zhou W, Xu F, Yuan S, Leng Y, Yuan Q (2018) Epileptic seizure prediction based on local mean decomposition and deep convolutional neural network. J Supercomput 76(5):3462–3476. https://doi.org/10.1007/s11227-018-2600-6

Yuan S, Zhou W, Chen L (2018) Epileptic seizure prediction using diffusion distance and Bayesian linear discriminate analysis on intracranial EEG. Int J Neural Syst 28(1):1750043. https://doi.org/10.1142/s0129065717500435

Zelig D, Goldberg I, Shor O, Ben Dor S, Yaniv-Rosenfeld A, Milikovsky DZ, Ofer J, Imtiaz H, Friedman A, Benninger F (2022) Paroxysmal slow wave events predict epilepsy following a first seizure. Epilepsia 63(1):190–198. https://doi.org/10.1111/epi.17110

Zhang Y, Yang S, Liu Y, Zhang Y, Han B, Zhou F (2018) Integration of 24 feature types to accurately detect and predict seizures using scalp EEG signals. Sensors (Basel), 18(5). https://doi.org/10.3390/s18051372

Zheng Y, Wang G, Li K, Bao G, Wang J (2014) Epileptic seizure prediction using phase synchronization based on bivariate empirical mode decomposition. Clin Neurophysiol 125(6):1104–1111. https://doi.org/10.1016/j.clinph.2013.09.047

Zhu J, Lin H, Su C, Shen Q, Jiang D (2003) The roles of different components of EEGs for seizure prediction-wavelet energy evaluation. Acta Bioch Bioph Sin 01:73–78

Google Scholar  

Zurich (2018) The SWEC-ETHZ iEEG database and algorithms. http://ieeg-swez.ethz.ch/

Download references

Acknowledgements

This work was supported in part by the National Key Research and Development Program of China under Grant 2022YFF1202304; in part by the National Natural Science Foundation of China under Grant 62376190, Grant 81925020, and Grant 82001939; and in part by the Natural Science Foundation of Tianjin under Grant 22JCYBJC01430.

National Natural Science Foundation of China 62376190, Zhongpeng Wang, 81925020, Dong Ming, 82001939, Long Chen

Author information

Authors and affiliations.

Academy of Medical Engineering and Translational Medicine, Tianjin University, Tianjin, 300072, China

Zhongpeng Wang, Xiaoxin Song, Long Chen, Jinxiang Nan, Yulin Sun, Meijun Pang, Kuo Zhang, Xiuyun Liu & Dong Ming

Haihe Laboratory of Brain-Computer Interaction and Human-Machine Integration, Tianjin, 300392, China

Zhongpeng Wang, Long Chen, Meijun Pang, Kuo Zhang, Xiuyun Liu & Dong Ming

You can also search for this author in PubMed   Google Scholar

Contributions

All authors wrote the paper.

Corresponding authors

Correspondence to Long Chen or Xiuyun Liu .

Ethics declarations

Conflict of interest.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data Availability

No data was used for the research described in the article.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Wang, Z., Song, X., Chen, L. et al. Research progress of epileptic seizure prediction methods based on EEG. Cogn Neurodyn (2024). https://doi.org/10.1007/s11571-024-10109-w

Download citation

Received : 20 September 2023

Revised : 09 March 2024

Accepted : 14 March 2024

Published : 07 May 2024

DOI : https://doi.org/10.1007/s11571-024-10109-w

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Seizure prediction
• Intracranial EEG (iEEG)
• Feature extraction
• Find a journal
• Publish with us
• Track your research

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Review Article
• Published: 29 April 2024

Krüppel-like factors: potential roles in blood-brain barrier dysfunction and epileptogenesis

• Ana Beatriz Santos 1 ,
• Andreia Carona 1 , 2 ,
• Miren Ettcheto 3 , 4 , 5 , 6 ,
• Antoni Camins 3 , 4 , 5 , 6 ,
• Amílcar Falcão 1 , 2 ,
• Ana Fortuna 1 , 2 &
• Joana Bicker 1 , 2  

Acta Pharmacologica Sinica ( 2024 ) Cite this article

87 Accesses

Metrics details

Epilepsy is a chronic and debilitating neurological disorder, known for the occurrence of spontaneous and recurrent seizures. Despite the availability of antiseizure drugs, 30% of people with epilepsy experience uncontrolled seizures and drug resistance, evidencing that new therapeutic options are required. The process of epileptogenesis involves the development and expansion of tissue capable of generating spontaneous recurrent seizures, during which numerous events take place, namely blood-brain barrier (BBB) dysfunction, and neuroinflammation. The consequent cerebrovascular dysfunction results in a lower seizure threshold, seizure recurrence, and chronic epilepsy. This suggests that improving cerebrovascular health may interrupt the pathological cycle responsible for disease development and progression. Krüppel-like factors (KLFs) are a family of zinc-finger transcription factors, encountered in brain endothelial cells, glial cells, and neurons. KLFs are known to regulate vascular function and changes in their expression are associated with neuroinflammation and human diseases, including epilepsy. Hence, KLFs have demonstrated various roles in cerebrovascular dysfunction and epileptogenesis. This review critically discusses the purpose of KLFs in epileptogenic mechanisms and BBB dysfunction, as well as the potential of their pharmacological modulation as therapeutic approach for epilepsy treatment.

This is a preview of subscription content, access via your institution

Access options

Subscribe to this journal

Receive 12 print issues and online access

251,40 € per year

only 20,95 € per issue

Buy this article

• Purchase on Springer Link
• Instant access to full article PDF

Prices may be subject to local taxes which are calculated during checkout

Similar content being viewed by others

Antisense oligonucleotide therapeutic approach for Timothy syndrome

Brain endothelial GSDMD activation mediates inflammatory BBB breakdown

The blood–brain barrier: structure, regulation, and drug delivery

World Health Organization. Epilepsy: a public health imperative. Geneva: World Health Organization. 2019. p. 1–171.

World Health Organization. Improving the lives of people with epilepsy: a technical brief. Geneva: World Health Organization. 2022. p. 1–43.

Clossen BL, Reddy DS. Novel therapeutic approaches for disease-modification of epileptogenesis for curing epilepsy. Biochim Biophys Acta Mol Basis Dis. 2017;1863:1519–38.

Article   CAS   PubMed   Google Scholar  

Heinemann U. Epilepsy; Basic mechanisms. In: Aminoff MJ, Daroff RB, editors. Encyclopedia of the Neurological Sciences . Amsterdam: Elsevier; 2014. p. 93–96.

Fauser S, Tumani H. Epilepsy. Handb Clin Neurol. 2017;146:259–66.

Article   PubMed   Google Scholar  

Thijs RD, Surges R, O’Brien TJ, Sander JW. Epilepsy in adults. Lancet. 2019;393:689–701.

Pitkanen A, Lukasiuk K, Dudek E, Staley KJ. Epileptogenesis. Cold Spring Harb Perspect Med. 2015;5:1–18.

Article   CAS   Google Scholar  

Pitkänen A, Engel J. Past and present definitions of epileptogenesis and its biomarkers. Neurotherapeutics. 2014;11:231–41.

Article   PubMed   PubMed Central   Google Scholar  

Klein P, Tyrlikova I. No prevention or cure of epilepsy as yet. Neuropharmacology. 2020;168:107762.

Mukhtar I. Inflammatory and immune mechanisms underlying epileptogenesis and epilepsy: From pathogenesis to treatment target. Seizure. 2020;82:65–79.

Hunt RF, Boychuk JA, Smith BN. Neural circuit mechanisms of post–traumatic epilepsy. Front Cell Neurosci. 2013;7:89.

Reddy SD, Younus I, Sridhar V, Reddy DS. Neuroimaging biomarkers of experimental epileptogenesis and refractory epilepsy. Int J Mol Sci. 2019;20:1–23.

Article   Google Scholar  

Lukasiuk K. Epileptogenesis. In: Aminoff MJ, Daroff RB, editors. Encyclopedia of the Neurological Sciences . Amsterdam: Elsevier; 2014. p. 196–199.

Löscher W, Hirsch LJ, Schmidt D. The enigma of the latent period in the development of symptomatic acquired epilepsy—Traditional view versus new concepts. Epilepsy Behav. 2015;52:78–92.

Saito T, Saito Y, Sugai K, Nakagawa E, Komaki H, Okazaki T, et al. Late-onset epilepsy in children with acute febrile encephalopathy with prolonged convulsions: a clinical and encephalographic study. Brain Dev. 2013;35:531–9.

Łukawski K, Andres-Mach M, Czuczwar M, Łuszczki JJ, Kruszyński K, Czuczwar SJ. Mechanisms of epileptogenesis and preclinical approach to antiepileptogenic therapies. Pharmacol Rep. 2018;70:284–93.

Youn Y, Sung IK, Lee IG. The role of cytokines in seizures: interleukin (IL)-1β, IL-1Ra, IL-8, and IL-10. Korean J Pediatr. 2013;56:271.

Article   CAS   PubMed   PubMed Central   Google Scholar  

Reiss Y, Bauer S, David B, Devraj K, Fidan E, Hattingen E, et al. The neurovasculature as a target in temporal lobe epilepsy. Brain Pathol. 2023;33:e13147.

Guo D, Zhang B, Han L, Rensing NR, Wong M. Cerebral vascular and blood brain–barrier abnormalities in a mouse model of epilepsy and tuberous sclerosis complex. Epilepsia. 2024;65:483–96.

Löscher W, Friedman A. Structural, molecular, and functional alterations of the blood-brain barrier during epileptogenesis and epilepsy: a cause, consequence, or both? Int J Mol Sci. 2020;21:591.

Obermeier B, Daneman R, Ransohoff RM. Development, maintenance and disruption of the blood-brain barrier. Nat Med. 2013;19:1584–96.

Swissa E, Serlin Y, Vazana U, Prager O, Friedman A. Blood-brain barrier dysfunction in status epileptics: mechanisms and role in epileptogenesis. Epilepsy Behav. 2019;101:106285.

van Lanen RHGJ, Haeren RHL, Staals J, Dings JTA, Schijns OEMG, Hoogland G, et al. Cerebrovascular glycocalyx damage and microcirculation impairment in patients with temporal lobe epilepsy. J Cereb Blood Flow Metab. 2023;43:1737–51.

Zhao F, Zhong L, Luo Y. Endothelial glycocalyx as an important factor in composition of blood-brain barrier. CNS Neurosci Ther. 2021;27:26–35.

Kutuzov N, Flyvbjerg H, Lauritzen M. Contributions of the glycocalyx, endothelium, and extravascular compartment to the blood–brain Barrier. Proc Natl Acad Sci USA. 2018;115:E9429–38.

Zhu J, Li X, Yin J, Hu Y, Gu Y, Pan S. Glycocalyx degradation leads to blood-brain barrier dysfunction and brain edema after asphyxia cardiac arrest in rats. J Cereb Blood Flow Metab. 2018;38:1979–92.

Takata F, Nakagawa S, Matsumoto J, Dohgu S. Blood-brain barrier dysfunction amplifies the development of neuroinflammation: understanding of cellular events in brain microvascular endothelial cells for prevention and treatment of BBB dysfunction. Front Cell Neurosci. 2021;15:661838.

van Vliet EA, Aronica E, Gorter JA. Role of blood-brain barrier in temporal lobe epilepsy and pharmacoresistance. Neuroscience. 2014;277:455–73.

Gilad R, Lampl Y, Eilam A, Boaz M, Loyberboim M. SPECT-DTPA as a tool for evaluating the blood-brain barrier in post-stroke seizures. J Neurol. 2012;259:2041–4.

Mendes NF, Pansani AP, Carmanhães ERF, Tange P, Meireles JV, Ochikubo M, et al. The blood-brain barrier breakdown during acute phase of the pilocarpine model of epilepsy Is dynamic and time-dependent. Front Neurol. 2019;10:382.

Liu WY, Wang ZB, Wang Y, Tong LC, Li Y, Wei X, et al. Increasing the permeability of the blood-brain barrier in three different models in vivo. CNS Neurosci Ther. 2015;21:568–74.

Morris G, Fernandes BS, Puri BK, Walker AJ, Carvalho AF, Berk M. Leaky brain in neurological and psychiatric disorders: drivers and consequences. Aust N Z J Psychiatry. 2018;52:924–48.

Greene C, Hanley N, Campbell M. Blood-brain barrier associated tight junction disruption is a hallmark feature of major psychiatric disorders. Transl Psychiatry. 2020;10:373.

Deshwar AR, Cytrynbaum C, Murthy H, Zon J, Chitayat D, Volpatti J, et al. Variants in CLDN5 cause a syndrome characterized by seizures, microcephaly and brain calcifications. Brain. 2023;146:2285–97.

Han H, Mann A, Ekstein D, Eyal S. Breaking bad: the structure and function of the blood-brain barrier in epilepsy. AAPS J. 2017;19:973–88.

Jiruska P, Freestone D, Gnatkovsky V, Wang Y. An update on the seizures beget seizures theory. Epilepsia. 2023;64:S13–S24.

Boison D, Sandau US, Ruskin DN, Kawamura M, Masino SA. Homeostatic control of brain function—new approaches to understand epileptogenesis. Front Cell Neurosci. 2013;7:109.

Sano F, Shigetomi E, Shinozaki Y, Tsuzukiyama H, Saito K, Mikoshiba K, et al. Reactive astrocyte-driven epileptogenesis is induced by microglia initially activated following status epilepticus. JCI Insight. 2021;6:e135391.

Benson MJ, Manzanero S, Borges K. Complex alterations in microglial M1/M2 markers during the development of epilepsy in two mouse models. Epilepsia. 2015;56:895–905.

Haruwaka K, Ikegami A, Tachibana Y, Ohno N, Konishi H, Hashimoto A, et al. Dual microglia effects on blood brain barrier permeability induced by systemic inflammation. Nat Commun. 2019;10:5816.

Henning L, Unichenko P, Bedner P, Steinhäuser C, Henneberger C. Astrocytes as initiators of epilepsy. Neurochem Res. 2022;48:1091–9.

Liddelow SA, Guttenplan KA, Clarke LE, Bennett FC, Bohlen CJ, Schirmer L, et al. Neurotoxic reactive astrocytes are induced by activated microglia. Nature 2017;541:481–7.

Ortinski PI, Dong J, Mungenast A, Yue C, Takano H, Watson DJ, et al. Selective induction of astrocytic gliosis generates deficits in neuronal inhibition. Nat Neurosci. 2010;13:584–91.

Gorter JA, Aronica E, van Vliet EA. The roof is leaking and a storm is raging: repairing the blood-brain barrier in the fight against epilepsy. Epilepsy Curr. 2019;19:177–81.

Zhang X, Wang D, Zhang B, Zhu J, Zhou Z, Cui L. Regulation of microglia by glutamate and its signal pathway in neurodegenerative diseases. Drug Discov Today. 2020;25:1074–85.

Librizzi L, Noè F, Vezzani A, De Curtis M, Ravizza T. Seizure-induced brain-borne inflammation sustains seizure recurrence and blood-brain barrier damage. Ann Neurol. 2012;72:82–90.

Marchi N, Granata T, Ghosh C, Janigro D. Blood-brain barrier dysfunction and epilepsy: pathophysiologic role and therapeutic approaches. Epilepsia. 2012;53:1877–86.

Luo J. TGF-β as a key modulator of astrocyte reactivity: disease relevance and therapeutic implications. Biomedicines. 2022;10:1206.

Diniz LP, Matias I, Siqueira M, Stipursky J, Gomes FCA. Astrocytes and the TGF-β1 pathway in the healthy and diseased brain: a double-edged sword. Mol Neurobiol. 2019;56:4653–79.

Weissberg I, Wood L, Kamintsky L, Vazquez O, Milikovsky DZ, Alexander A, et al. Albumin induces excitatory synaptogenesis through astrocytic TGF-β/ALK5 signaling in a model of acquired epilepsy following blood-brain barrier dysfunction. Neurobiol Dis. 2015;78:115–25.

Michalak Z, Sano T, Engel T, Miller-Delaney SFC, Lerner-Natoli M, Henshall DC. Spatio-temporally restricted blood-brain barrier disruption after intra-amygdala kainic acid-induced status epilepticus in mice. Epilepsy Res. 2013;103:167–79.

Van Vliet EA, Otte WM, Wadman WJ, Aronica E, Kooij G, De Vries HE, et al. Blood-brain barrier leakage after status epilepticus in rapamycin-treated rats I: magnetic resonance imaging. Epilepsia. 2016;57:59–69.

Maroso M, Balosso S, Ravizza T, Liu J, Aronica E, Iyer AM, et al. Toll-like receptor 4 and high-mobility group box-1 are involved in ictogenesis and can be targeted to reduce seizures. Nat Med. 2010;16:413–9.

Patterson KP, Brennan GP, Curran M, Kinney-Lang E, Dubé C, Rashid F, et al. Rapid, coordinate inflammatory responses after experimental febrile status epilepticus: implications for epileptogenesis. eNeuro. 2015;2:0034–15.

Ravizza T, Vezzani A. Pharmacological targeting of brain inflammation in epilepsy: therapeutic perspectives from experimental and clinical studies. Epilepsia Open. 2018;3:133–42.

de Vries EE, van den Munckhof B, Braun KPJ, van Royen-Kerkhof A, de Jager W, Jansen FE. Inflammatory mediators in human epilepsy: a systematic review and meta-analysis. Neurosci Biobehav Rev. 2016;63:177–90.

Sweet DR, Fan L, Hsieh PN, Jain MK. Krüppel-like factors in vascular inflammation: mechanistic insights and therapeutic potential. Front Cardiovasc Med. 2018;5:6.

Pollak NM, Hoffman M, Goldberg IJ, Drosatos K. Krüppel-like factors: crippling and un-crippling metabolic pathways. JACC Basic Transl Sci. 2018;3:132–56.

Tang X, Liu K, Hamblin MH, Xu Y, Yin KJ. Genetic deletion of krüppel-Like factor 11 aggravates ischemic brain injury. Mol Neurobiol. 2018;55:2911–21.

Yin KJ, Hamblin M, Fan Y, Zhang J, Chen YE. Krüpple-like factors in the central nervous system: novel mediators in stroke. Metab Brain Dis. 2015;30:401–10.

Zhou C, Sun P, Hamblin MH, Yin KJ. Genetic deletion of Krüppel-like factor 11 aggravates traumatic brain injury. J Neuroinflammation. 2022;19:281.

McConnell BB, Yang VW. Mammalian Krüppel-like factors in health and diseases. Physiol Rev. 2010;90:1337–81.

Kotlyarov S, Kotlyarova A. Participation of Krüppel-like factors in atherogenesis. Metabolites. 2023;13:448.

Jha P, Das H. KLF2 in regulation of NF-κB-mediated immune cell function and inflammation. Int J Mol Sci. 2017;18:2383.

Presnell JS, Schnitzler CE, Browne WE. KLF/SP transcription factor family evolution: expansion, diversification, and innovation in eukaryotes. Genome Biol Evol. 2015;7:2289–309.

Kaushik DK, Gupta M, Das S, Basu A. Krüppel-like factor 4, a novel transcription factor regulates microglial activation and subsequent neuroinflammation. J Neuroinflammation. 2010;7:68.

Turpaev KT. Transcription factor KLF2 and its role in the regulation of inflammatory processes. Biochemistry. 2020;85:54–67.

CAS   PubMed   Google Scholar  

Ogryzko VV, Schiltz RL, Russanova V, Howard BH, Nakatani Y. The transcriptional coactivators p300 and CBP are histone acetyltransferases. Cell. 1996;87:953–9.

Pons D, De Vries FR, Van Den Elsen PJ, Heijmans BT, Quax PHA, Jukema JW. Epigenetic histone acetylation modifiers in vascular remodelling: new targets for therapy in cardiovascular disease. Eur Heart J. 2009;30:266–77.

Chen Y, Chen J, Chen Y, Li Y. miR-146a/KLF4 axis in epileptic mice: a novel regulator of synaptic plasticity involving STAT3 signaling. Brain Res. 2022;1790:147988.

Thompson R, Chan C. NRSF and its epigenetic effectors: new treatments for neurological disease. Brain Sci. 2018;8:1–12.

Jin L, Liu Y, Wu Y, Huang Y, Zhang D. REST is not resting: REST/NRSF in health and disease. Biomolecules. 2023;13:1–23.

Carminati E, Buffolo F, Rocchi A, Michetti C, Cesca F, Benfenati F. Mild inactivation of RE-1 Silencing Transcription Factor (REST) reduces susceptibility to kainic acid-induced seizures. Front Cell Neurosci. 2020;13:1–10.

Navarrete-Modesto V, Orozco-Suárez S, Alonso-Vanegas M, Feria-Romero IA, Rocha L. REST/NRSF transcription factor is overexpressed in hippocampus of patients with drug-resistant mesial temporal lobe epilepsy. Epilepsy Behav. 2019;94:118–23.

McClelland S, Brennan GP, Dubé C, Rajpara S, Iyer S, Richichi C, et al. The transcription factor NRSF contributes to epileptogenesis by selective repression of a subset of target genes. Elife. 2014;3:e01267.

Natali G, Michetti C, Krawczun-Rygmaczewska A, Floss T, Cesca F, Benfenati F. Conditional knockout of REST/NRSF in excitatory neurons reduces seizure susceptibility to chemical kindling. Front Cell Neurosci. 2023;17:1267609.

Prada I, Marchaland J, Podini P, Magrassi L, D’Alessandro R, Bezzi P, et al. REST/NRSF governs the expression of dense-core vesicle gliosecretion in astrocytes. J Cell Biol. 2011;193:537–49.

Schiffer D, Caldera V, Mellai M, Conforti P, Cattaneo E, Zuccato C. Repressor element-1 silencing transcription factor (REST) is present in human control and Huntington’s disease neurones. Neuropathol Appl Neurobiol. 2014;40:899–910.

Palm K, Belluardo N, Metsis M, Timmusk T. Neuronal expression of zinc finger transcription factor REST/NRSF/XBR gene. J Neurosci. 1998;18:1280–96.

Parrish JZ, Kim CC, Tang L, Bergquist S, Wang T, DeRisi JL, et al. Krüppel mediates the selective rebalancing of ion channel expression. Neuron. 2014;82:537–44.

Jeong KH, Lee KE, Kim SY, Cho KO. Upregulation of Krüppel-like factor 6 in the mouse hippocampus after pilocarpine-induced status epilepticus. Neuroscience. 2011;186:170–8.

Orlova KA, Tsai V, Baybis M, Heuer GG, Sisodiya S, Thom M, et al. Early progenitor cell marker expression distinguishes type II from type I focal cortical dysplasias. J Neuropathol Exp Neurol. 2010;69:850–63.

Yao K, Duan Z, Zhou J, Li L, Zhai F, Dong Y, et al. Clinical and immunohistochemical characteristics of type II and type I focal cortical dysplasia. Oncotarget. 2016;7:76415–22.

Zhu S, Tai C, MacVicar BA, Jia W, Cynader MS. Glutamatergic stimulation triggers rapid Krüpple-like factor 4 expression in neurons and the overexpression of KLF4 sensitizes neurons to NMDA-induced caspase-3 activity. Brain Res. 2009;1250:49–62.

Zhou S, Li J, Zhang XN, Xiong W. MicroRNA-124 modulates neuroinflammation in acute methanol poisoning rats via targeting Krüppel-like factor-6. Bioengineered. 2022;13:13507–19.

Ávila-Mendoza J, Subramani A, Sifuentes CJ, Denver RJ. Molecular mechanisms for Krüppel-like Factor 13 actions in hippocampal neurons. Mol Neurobiol. 2020;57:3785–802.

Tapia-Ramírez J, Eggen BJL, Peral-Rubio MJ, Toledo-Aral JJ, Mandel G. A single zinc finger motif in the silencing factor REST represses the neural-specific type II sodium channel promoter. Proc Natl Acad Sci USA. 1997;94:1177–82.

Yeo M, Berglund K, Augustine G, Liedtke W. Novel repression of Kcc2 transcription by REST-RE-1 controls developmental switch in neuronal chloride. J Neurosci. 2009;29:14652–62.

McMoneagle E, Zhou J, Zhang S, Huang W, Josiah SS, Ding K, et al. Neuronal K +  -Cl - cotransporter KCC2 as a promising drug target for epilepsy treatment. Acta Pharmacol Sin. 2024;45:1–22.

Spencer EM, Chandler KE, Haddley K, Howard MR, Hughes D, Belyaev ND, et al. Regulation and role of REST and REST4 variants in modulation of gene expression in in vivo and in vitro in epilepsy models. Neurobiol Dis. 2006;24:41–52.

McClelland S, Flynn C, Dubé C, Richichi C, Zha Q, Ghestem A, et al. Neuron-restrictive silencer factor-mediated hyperpolarization-activated cyclic nucleotide gated channelopathy in experimental temporal lobe epilepsy. Ann Neurol. 2011;70:454–65.

Hu XL, Cheng X, Cai L, Tan GH, Xu L, Feng XY, et al. Conditional deletion of NRSF in forebrain neurons accelerates epileptogenesis in the kindling model. Cereb Cortex. 2011;21:2158–65.

Aronica E, Bauer S, Bozzi Y, Caleo M, Dingledine R, Gorter JA, et al. Neuroinflammatory targets and treatments for epilepsy validated in experimental models. Epilepsia. 2017;58:27–38.

Zhang X, Tang X, Ma F, Fan Y, Sun P, Zhu T, et al. Endothelium-targeted overexpression of Krüppel-like factor 11 protects the blood-brain barrier function after ischemic brain injury. Brain Pathol. 2020;30:746–65.

Nayak L, Goduni L, Takami Y, Sharma N, Kapil P, Jain MK, et al. Kruppel-like factor 2 is a transcriptional regulator of chronic and acute inflammation. Am J Pathol. 2013;182:1696–704.

SenBanerjee S, Lin Z, Atkins GB, Greif DM, Rao RM, Kumar A, et al. KLF2 Is a novel transcriptional regulator of endothelial proinflammatory activation. J Exp Med. 2004;199:1305–15.

Liao X, Sharma N, Kapadia F, Zhou G, Lu Y, Hong H, et al. Krüppel-like factor 4 regulates macrophage polarization. J Clin Invest. 2011;121:2736–49.

Lin Z, Natesan V, Shi H, Dong F, Kawanami D, Mahabeleshwar GH, et al. Kruppel-like factor 2 regulates endothelial barrier function. Arterioscler Thromb Vasc Biol. 2010;30:1952–9.

Shi H, Sheng B, Zhang F, Wu C, Zhang R, Zhu J, et al. Kruppel-like factor 2 protects against ischemic stroke by regulating endothelial blood brain barrier function. Am J Physiol Heart Circ Physiol. 2013;304:796–805.

Park JH, Riew TR, Shin YJ, Park JM, Cho JM, Lee MY. Induction of Krüppel-like factor 4 expression in reactive astrocytes following ischemic injury in vitro and in vivo. Histochem Cell Biol. 2014;141:33–42.

Wang C, Li L. The critical role of KLF4 in regulating the activation of A1/A2 reactive astrocytes following ischemic stroke. J Neuroinflammation. 2023;20:40.

Zhang X, Wang L, Han Z, Dong J, Pang D, Fu Y, et al. KLF4 alleviates cerebral vascular injury by ameliorating vascular endothelial inflammation and regulating tight junction protein expression following ischemic stroke. J Neuroinflammation. 2020;17:107.

Ma J, Wang P, Liu Y, Zhao L, Li Z, Xue Y. Krüppel-like factor 4 regulates blood-tumor barrier permeability via ZO-1, occludin and claudin-5. J Cell Physiol. 2014;229:916–26.

Cowan CE, Kohler EE, Dugan TA, Mirza MK, Malik AB, Wary KK. Kruppel-like factor-4 transcriptionally regulates VE-cadherin expression and endothelial barrier function. Circ Res. 2010;107:959–66.

Wen M, Ye J, Han Y, Huang L, Yang H, Jiang W, et al. Hypertonic saline regulates microglial M2 polarization via miR-200b/KLF4 in cerebral edema treatment. Biochem Biophys Res Commun. 2018;499:345–53.

Wang X, Li H, Chen S, He J, Chen W, Ding Y, et al. P300/CBP-associated factor (PCAF) attenuated M1 macrophage inflammatory responses possibly through KLF2 and KLF4. Immunol Cell Biol. 2021;99:724–36.

El-Deeb NK, El-Tanbouly DM, Khattab MA, EL-Yamany MF, Mohamed AF. Crosstalk between PI3K/AKT/KLF4 signaling and microglia M1/M2 polarization as a novel mechanistic approach towards flibanserin repositioning in Parkinson’s disease. Int Immunopharmacol. 2022;112:109191.

Kapoor N, Niu J, Saad Y, Kumar S, Sirakova T, Becerra E, et al. Transcription factors STAT6 and KLF4 implement macrophage polarization via the dual catalytic powers of MCPIP. J Immunol. 2015;194:6011–23.

Ghosh A, Pahan K. Gemfibrozil, a lipid-lowering drug, induces suppressor of cytokine signaling 3 in glial cells: implications for neurodegenerative disorders. J Biol Chem. 2012;287:27189–203.

Kaushik DK, Mukhopadhyay R, Kumawat KL, Gupta M, Basu A. Therapeutic targeting of Krüppel-like factor 4 abrogates microglial activation. J Neuroinflammation. 2012;9:57.

Kaushik DK, Thounaojam MC, Kumawat KL, Gupta M, Basu A. Interleukin-1β orchestrates underlying inflammatory responses in microglia via Krüppel-like factor 4. J Neurochem. 2013;127:233–44.

Uta R, François C, Marvin H, Philipp A, Andrea PP, Henrik W, et al. Anti-inflammatory properties of Honokiol in activated primary microglia and astrocytes. J Neuroimmunol. 2018;323:78–86.

Zhuo Z, Wang Y, Kong H, Fu T. GKLF, a transcriptional activator of Txnip, drives microglia activation in kainic acid-induced murine models of epileptic seizures. Int Immunopharmacol. 2023;121:110426.

Evans PM, Zhang W, Chen X, Yang J, Bhakat KK, Liu C. Kruppel-like factor 4 is acetylated by p300 and regulates gene transcription via modulation of histone acetylation. J Biol Chem. 2007;282:33994–4002.

Zhang Y, Lei CQ, Hu YH, Xia T, Li M, Zhong B, et al. Krüppel-like factor 6 is a co-activator of NF-κB that mediates p65-dependent transcription of selected downstream genes. J Biol Chem. 2014;289:12876–85.

Buffolo F, Petrosino V, Albini M, Moschetta M, Carlini F, Floss T, et al. Neuroinflammation induces synaptic scaling through IL-1β-mediated activation of the transcriptional repressor REST/NRSF. Cell Death Dis. 2021;12:180.

Wang W, Wang X, Chen L, Zhang Y, Xu Z, Liu J, et al. The microRNA miR-124 suppresses seizure activity and regulates CREB1 activity. Expert Rev Mol Med. 2016;18:e4.

Tuomisto TT, Lumivuori H, Kansanen E, Häkkinen SK, Turunen MP, Van Thienen JV, et al. Simvastatin has an anti-inflammatory effect on macrophages via upregulation of an atheroprotective transcription factor, Kruppel-like factor 2. Cardiovasc Res. 2008;78:175–84.

Maejima T, Inoue T, Kanki Y, Kohro T, Li G, Ohta Y, et al. Direct evidence for pitavastatin induced chromatin structure change in the KLF4 gene in endothelial cells. PLoS One. 2014;9:e96005.

Ohnesorge N, Viemann D, Schmidt N, Czymai T, Spiering D, Schmolke M, et al. Erk5 activation elicits a vasoprotective endothelial phenotype via induction of Kruppel-like factor 4 (KLF4). J Biol Chem. 2010;285:26199–210.

Bu DX, Tarrio M, Grabie N, Zhang Y, Yamazaki H, Stavrakis G, et al. Statin-induced Krüppel-like factor 2 expression in human and mouse T cells reduces inflammatory and pathogenic responses. J Clin Invest. 2010;120:1961–70.

Sen-Banerjee S, Mir S, Lin Z, Hamik A, Atkins GB, Das H, et al. Kruppel-like factor 2 as a novel mediator of statin effects in endothelial cells. Circulation. 2005;112:720–6.

Wang Y, Zhao B, Zhang Y, Tang Z, Shen Q, Zhang Y, et al. Krüppel-like factor 4 is induced by rapamycin and mediates the anti-proliferative effect of rapamycin in rat carotid arteries after balloon injury. Br J Pharmacol. 2012;165:2378–88.

Chen S, Ye J, Chen X, Shi J, Wu W, Lin W, et al. Valproic acid attenuates traumatic spinal cord injury-induced inflammation via STAT1 and NF-κB pathway dependent of HDAC3. J Neuroinflammation. 2018;15:150.

Wang Z, Leng Y, Tsai LK, Leeds P, Chuang DM. Valproic acid attenuates blood-brain barrier disruption in a rat model of transient focal cerebral ischemia: the roles of HDAC and MMP-9 inhibition. J Cereb Blood Flow Metab. 2011;31:52–7.

Gong F, Shi Q, Mou X, Wang K, Wang Q, Wang H. Atorvastatin mitigates memory deficits and brain monocyte infiltration in chronic hypercholesterolemia. Aging 2023;15:13669–79.

Parmar KM, Nambudiri V, Dai G, Larman HB, Gimbrone MA, García-Cardeña G. Statins exert endothelial atheroprotective effects via the KLF2 transcription factor. J Biol Chem. 2005;280:26714–9.

Guo J, Li J, Zhou M, Qin F, Zhang S, Wu B, et al. Statin treatment reduces the risk of poststroke seizures. Neurology 2015;85:701–7.

Hufthy Y, Bharadwaj M, Gupta S, Hussain D, Joseph PJS, Khan A, et al. Statins as antiepileptogenic drugs: analyzing the evidence and identifying the most promising statin. Epilepsia 2022;63:1889–98.

Etminan M, Samii A, Brophy JM. Statin use and risk of epilepsy: a nested case-control study. Neurology 2010;75:1496–500.

Lin FJ, Lin HW, Ho YF. Effect of statin intensity on the risk of epilepsy after ischaemic stroke: real-world evidence from population-based health claims. CNS Drugs. 2018;32:367–76.

de Oliveira CV, Grigoletto J, Canzian JM, Duarte MMMF, Duarte T, Furian AF, et al. Effect of atorvastatin on behavioral alterations and neuroinflammation during epileptogenesis. Epilepsy Behav. 2018;78:109–17.

Üzüm G, Akgün-Dar K, Aksu U. The effects of atorvastatin on memory deficit and seizure susceptibility in pentylentetrazole-kindled rats. Epilepsy Behav. 2010;19:284–9.

Iqubal A, Sharma S, Sharma K, Bhavsar A, Hussain I, Iqubal MK, et al. Intranasally administered pitavastatin ameliorates pentylenetetrazol-induced neuroinflammation, oxidative stress and cognitive dysfunction. Life Sci. 2018;211:172–81.

Hamik A, Lin Z, Kumar A, Balcells M, Sinha S, Katz J, et al. Kruppel-like factor 4 regulates endothelial inflammation. J Biol Chem. 2007;282:13769–79.

Curatolo P, Moavero R. mTOR inhibitors as a new therapeutic option for epilepsy. Expert Rev Neurother. 2013;13:627–38.

Van Vliet EA, Forte G, Holtman L, Den Burger JCG, Sinjewel A, De Vries HE, et al. Inhibition of mammalian target of rapamycin reduces epileptogenesis and blood-brain barrier leakage but not microglia activation. Epilepsia 2012;53:1254–63.

Kwon IS, Wang W, Xu S, Jin ZG. Histone deacetylase 5 interacts with Krüppel-like factor 2 and inhibits its transcriptional activity in endothelium. Cardiovasc Res. 2014;104:127–37.

Zohre S, Kazem NK, Abolfazl A, Mohammad RY, Aliakbar M, Alizadeh E, et al. Trichostatin A-induced apoptosis is mediated by Kruppel-like factor 4 in ovarian and lung cancer. Asian Pac J Cancer Prev. 2014;15:6581–6.

Yu T, Chen X, Zhang W, Liu J, Avdiushko R, Napier DL, et al. KLF4 regulates adult lung tumor-initiating cells and represses K-Ras-mediated lung cancer. Cell Death Differ. 2016;23:207–15.

Bicker J, Alves G, Fonseca C, Falcão A, Fortuna A. Repairing blood-CNS barriers: future therapeutic approaches for neuropsychiatric disorders. Pharm Res. 2020;162:105226.

Moschidou D, Mukherjee S, Blundell MP, Drews K, Jones GN, Abdulrazzak H, et al. Valproic acid confers functional pluripotency to human amniotic fluid stem cells in a transgene-free approach. Mol Ther. 2012;20:1953–67.

Kee HJ, Kook H. Krüppel-like factor 4 mediates histone deacetylase inhibitor-induced prevention of cardiac hypertrophy. J Mol Cell Cardiol. 2009;47:770–80.

Kim SJ, Lee BH, Lee YS, Kang KS. Defective cholesterol traffic and neuronal differentiation in neural stem cells of Niemann-Pick type C disease improved by valproic acid, a histone deacetylase inhibitor. Biochem Biophys Res Commun. 2007;360:593–9.

van Vliet EA, Marchi N. Neurovascular unit dysfunction as a mechanism of seizures and epilepsy during aging. Epilepsia 2022;63:1297–313.

Rempe RG, Hartz AMS, Soldner ELB, Sokola BS, Alluri SR, Abner EL, et al. Matrix metalloproteinase-mediated blood-brain barrier dysfunction in epilepsy. J Neurosci. 2018;38:4301–15.

Download references

Acknowledgements

The authors acknowledge Fundo Europeu de Desenvolvimento Regional (FEDER) through Portugal2020 in the scope of the Operational Programme for Competitiveness and Internationalization; Fundação para a Ciência e Tecnologia (FCT), Portuguese Agency for Scientific Research, within the scope of the research project 2022.03133.PTDC and Ph.D. research grant 2021.06125.BD; Spanish Ministerio de Ciencia e Innovación (PID2021-123462OB-I00); the Generalitat de Catalunya (2021 SGR 00288); CIBERNED (Grant CB06/05/2004); and Institut de Neurociències UB, (CEX2021-001159-M). ME is Serra Hunter fellow.

Author information

Authors and affiliations.

University of Coimbra, Faculty of Pharmacy, Coimbra, Portugal

Ana Beatriz Santos, Andreia Carona, Amílcar Falcão, Ana Fortuna & Joana Bicker

University of Coimbra, Coimbra Institute for Biomedical Imaging and Translational Research, Coimbra, Portugal

Andreia Carona, Amílcar Falcão, Ana Fortuna & Joana Bicker

Biomedical Research Networking Center in Neurodegenerative Diseases (CIBERNED), Madrid, Spain

Miren Ettcheto & Antoni Camins

Department of Pharmacology, Toxicology and Therapeutic Chemistry, Faculty of Pharmacy and Food Science, Universitat de Barcelona, Barcelona, Spain

Institute of Neuroscience, Universitat de Barcelona, Barcelona, Spain

Institut d’Investigació Sanitària Pere Virgili (IISPV), Reus, Spain

You can also search for this author in PubMed   Google Scholar

Contributions

Ana Beatriz Santos: Conceptualization, Data curation, Formal analysis, Investigation, Writing—original draft, Writing—review & editing; Andreia Carona: Formal analysis, Writing—original draft, Writing—review & editing; Miren Ettcheto: Formal analysis, Writing—review & editing; Antoni Camins: Formal analysis, Writing—review & editing; Amílcar Falcão: Formal analysis, Writing—review & editing; Ana Fortuna: Data curation, Formal analysis, Writing—original draft, Writing—review & editing; Joana Bicker: Conceptualization; Formal analysis, Writing—original draft, Writing—review & editing; Funding acquisition.

Corresponding author

Correspondence to Joana Bicker .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Supplementary information

Supplementary material, rights and permissions.

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Cite this article.

Santos, A.B., Carona, A., Ettcheto, M. et al. Krüppel-like factors: potential roles in blood-brain barrier dysfunction and epileptogenesis. Acta Pharmacol Sin (2024). https://doi.org/10.1038/s41401-024-01285-w

Download citation

Received : 22 December 2023

Accepted : 07 April 2024

Published : 29 April 2024

DOI : https://doi.org/10.1038/s41401-024-01285-w

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• blood-brain barrier
• cerebrovascular dysfunction
• epigenetics
• krüppel-like factors
• neuroinflammation

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

The Epidemiology of Epilepsy

Affiliation.

• 1 Department of Neuroscience, Istituto di Ricerche Farmacologiche Mario Negri IRCCS, Milan, Italy, [email protected].
• PMID: 31852003
• DOI: 10.1159/000503831

Epilepsy is a chronic disease of the brain characterized by an enduring (i.e., persisting) predisposition to generate seizures, unprovoked by any immediate central nervous system insult, and by the neurobiologic, cognitive, psychological, and social consequences of seizure recurrences. Epilepsy affects both sexes and all ages with worldwide distribution. The prevalence and the incidence of epilepsy are slightly higher in men compared to women and tend to peak in the elderly, reflecting the higher frequency of stroke, neurodegenerative diseases, and tumors in this age-group. Focal seizures are more common than generalized seizures both in children and in adults. The etiology of epilepsy varies according to the sociodemographic characteristics of the affected populations and the extent of the diagnostic workup, but a documented cause is still lacking in about 50% of cases from high-income countries (HIC). The overall prognosis of epilepsy is favorable in the majority of patients when measured by seizure freedom. Reports from low/middle-income countries (LMIC; where patients with epilepsy are largely untreated) give prevalence and remission rates that overlap those of HICs. As the incidence of epilepsy appears higher in most LMICs, the overlapping prevalence can be explained by misdiagnosis, acute symptomatic seizures and premature mortality. Studies have consistently shown that about one-half of cases tend to achieve prolonged seizure remission. However, more recent reports on the long-term prognosis of epilepsy have identified differing prognostic patterns, including early and late remission, a relapsing-remitting course, and even a worsening course (characterized by remission followed by relapse and unremitting seizures). Epilepsy per se carries a low mortality risk, but significant differences in mortality rates are expected when comparing incidence and prevalence studies, children and adults, and persons with idiopathic and symptomatic seizures. Sudden unexplained death is most frequent in people with generalized tonic-clonic seizures, nocturnal seizures, and drug refractory epilepsy.

Keywords: Burden; Epilepsy; Incidence; Mortality; Prevalence.

© 2019 S. Karger AG, Basel.

Publication types

• Epilepsy / epidemiology*
• Epilepsy / etiology
• Epilepsy / mortality
• Sudden Unexpected Death in Epilepsy*

Epilepsy and Seizures

What are epilepsy and seizures.

Epilepsy is a chronic brain disorder in which groups of nerve cells, or neurons, in the brain sometimes send the wrong signals and cause seizures. Neurons normally generate electrical and chemical signals that act on other neurons, organs, and muscles to produce human thoughts, feelings, and actions.

During a seizure, many neurons send signals at the same time, much faster than normal. This surge of excessive electrical activity may cause involuntary movements, sensations, emotions, and/or behaviors. The disturbance of normal nerve cell activity may cause a loss of awareness. Some people recover immediately after a seizure, while others may take minutes to hours to feel like themselves again. During this time, they may feel tired, sleepy, weak, or confused.

Epilepsy (sometimes referred to as a seizure disorder) can have many different causes and seizure types. Epilepsy varies in severity and impact from person to person and can be accompanied by a range of co-existing conditions. Epilepsy is sometimes called “the epilepsies” because of the diversity of types and causes. Some people may have convulsions (muscles contract repeatedly) and lose consciousness. Others may simply stop what they are doing, have a brief lapse of awareness, and stare into space for a short period. Some people have seizures very infrequently, while other people may experience hundreds of seizures each day.

While any seizure is cause for concern, having a seizure does not by itself mean a person has epilepsy. First seizures, febrile seizures, nonepileptic events, and eclampsia (a life-threatening condition that can occur during pregnancy) are examples of conditions involving seizures that may not be associated with epilepsy. Regardless of the type of seizure, it's important to inform your doctor when you have a seizure.

Who is more likely to have epilepsy and seizures?

Anyone can develop epilepsy. It affects both men and women of all races, ethnic backgrounds, and ages.

Epilepsy has many possible causes, but about half of people living with epilepsy do not know the cause. In some cases, epilepsy is clearly linked to genetic factors, developmental brain abnormalities, infection, traumatic brain injury (TBI) , stroke , brain tumors , or other identifiable problems. Anything that disturbs the normal pattern of nerve cell activity—from illness to brain damage to brain development problems—can lead to seizures.

Epilepsy may develop because of problems in the brain’s wiring, an imbalance of nerve signaling in the brain (in which some cells are unusually active or stop other brain cells from sending messages), or some combination of these factors. Sometimes, when the brain tries to repair itself after a head injury, stroke, or other problem, it can unintentionally create nerve connection issues that lead to seizures.

The role of genes in epilepsy

Genetic changes may play a key role in the development of certain types of epilepsy. Many types affect multiple members of a family, pointing to an inherited gene or genes. In other cases, gene variations may occur spontaneously and contribute to the development of epilepsy in people with no family history of the disorder (called “de novo” mutations). Overall, researchers estimate that hundreds of genes could play a role.

Several types of epilepsy (called channelopathy-associated epilepsy) have been linked to variations in genes that provide instructions for ion channels, the "gates" that control the flow of ions (charged molecules) in and out of cells and help regulate neuronal signaling. Other genetic changes that may play a role in epilepsy include variations in genes that control how neurons move through the brain during development (neuronal migration) and genes that help break down carbohydrates in the brain.

Other genetic changes may not cause epilepsy but may influence the disorder in other ways, for example, some genes may affect a person's susceptibility to seizures and responsiveness to anti-seizure medications.

Conditions that can lead to epilepsy

Epilepsy may develop as a result of many types of conditions that disrupt normal brain activity, known as “co-occurring conditions”. Once these conditions are treated, individuals may no longer have seizures. However, whether the seizures stop varies based on the type of disorder, the brain region that is affected, and how much brain damage occurred prior to treatment. Examples of conditions that can lead to epilepsy include:

• Brain tumors
• Head trauma
• Alcoholism or alcohol withdrawal
• Alzheimer's disease
• Strokes, heart attacks, and other conditions that deprive the brain of oxygen
• Abnormal blood vessel formation (called arteriovenous malformations) or bleeding in the brain
• Brain inflammation or swelling
• Infections such as meningitis, HIV-related infections, and viral encephalitis

Cerebral palsy and other developmental disorders also may be associated with epilepsy. Epilepsy is often seen in people with other brain development disorders, for example, among individuals with autism spectrum disorder or intellectual disabilities.

What can trigger seizures?

Seizure triggers do not cause epilepsy but can provoke seizures in those who are susceptible. For those who are already diagnosed with epilepsy and taking medication, triggers can spark a seizure. Triggers include:

• Drinking alcohol, or alcohol withdrawal
• Dehydration or missing meals
• Exposure to toxins or poisons, including lead, carbon monoxide, illicit drugs, and very large doses of prescription medications
• Hormonal changes associated with the menstrual cycle
• Sleep deprivation
• Visual stimulation such as flashing lights or moving patterns

In surveys of people with epilepsy, stress is the most commonly reported seizure trigger.

Types of seizures

Seizures are divided into two broad categories: focal seizures and generalized seizures. There are many different types of seizures within each of these categories.

Focal seizures

Focal seizures come from just one part of the brain. About 60% of people with epilepsy have focal seizures. These seizures are frequently described by the area of the brain in which they originate. For example, many people are diagnosed with focal frontal lobe or medial temporal lobe seizures.

Symptoms of focal seizures

In some focal seizures, the person remains conscious during the seizure but may experience motor, sensory, or psychic feelings (for example, intense dejà vu or memories) or sensations. The person may experience sudden and unexplainable feelings of joy, anger, sadness, or nausea. He or she also may hear, smell, taste, see, or feel things that are not real and may have movements of just one part of the body, for example, just one hand.

In other focal seizures, the person has a change in consciousness, which can produce a dreamlike experience. The person may display unusual, repetitive behaviors such as blinks, twitches, mouth movements, or even walking in a circle. These repetitive movements are called automatisms. A person may perform more complicated actions, which may seem purposeful, involuntarily. Individuals may also continue activities they started before the seizure began, such as washing dishes in a repetitive, unproductive fashion. These seizures usually last no more than a minute or two.

Some people with focal seizures may experience auras—unusual sensations that warn of an impending seizure. An individual's symptoms, and how they progress, tend to be similar every time. Some people report experiencing a prodrome, a feeling that a seizure is imminent lasting hours or days.

Following focal seizures, a person may experience symptoms in areas controlled by the area of the brain where their seizure originated. This can help doctors locate the brain region where the seizure started. After a seizure, some people may experience a headache or pain in the muscles that contracted during the seizure.

The symptoms of focal seizures can be easily confused with other disorders. The strange behavior and sensations caused by focal seizures also can be mistaken for symptoms of narcolepsy, fainting, or even mental illness. Several tests and careful monitoring may be needed to make the distinction between epilepsy and these other disorders.

Generalized seizures

Generalized seizures are a result of abnormal neuronal activity that rapidly emerges on both sides of the brain. These seizures may cause loss of consciousness, falls, or massive muscle contractions. Types of generalized seizures and their effects include:

• Absence seizures may cause the person to appear to be staring into space, with or without slight twitching of the muscles
• Tonic seizures  cause a stiffening of muscles of the body, generally in the back, legs, and arms
• Clonic seizures  cause repeated jerking movements of muscles on both sides of the body
• Myoclonic seizures  cause jerks or twitches of the upper body, arms, or legs
• Atonic seizures  cause a loss of normal muscle tone, which can cause the person to fall or drop the head involuntarily
• Tonic-clonic seizures  cause a combination of symptoms, including stiffening of the body and repeated jerks of the arms and/or legs as well as loss of consciousness
• Secondary generalized seizures begins in one part of the brain, then spreads to both halves of the brain (basically, a focal seizure followed by a generalized seizure)

Other types of seizures

Not all seizures can be easily defined as either focal or generalized. Some people have both types of seizures but with no clear pattern.

Febrile seizures are seizures that happen when a child has an illness that causes a high fever. Children who have febrile seizures are typically not prescribed antiseizure medications unless they have a family history of epilepsy, signs of nervous system impairment before the seizure, or have a relatively long or complicated seizure or more than one febrile seizure. The risk of subsequent non-febrile seizures is low unless one of these factors is present.

First seizures affect many people who have a single seizure at some point in their lives. It can be provoked or unprovoked, meaning that they can occur with or without any obvious triggering factors. Unless the person has suffered brain damage or there is a family history of epilepsy or other neurological abnormalities, most single seizures usually are not followed by additional seizures. Medical disorders which can provoke a seizure include:

• Low or very high blood sugar
• Changes in chemical levels in the blood (sodium, calcium, magnesium)
• Eclampsia during or after pregnancy
• Impaired function of the kidneys or liver

In some cases where additional epilepsy risk factors are present, drug treatment after the first seizure may help prevent future seizures. Evidence suggests that it may be beneficial to begin antiseizure medication once a person has had a second unprovoked seizure, as the chance of future seizures increases significantly after this occurs. A person with a pre-existing brain problem, for example, a prior stroke or traumatic brain injury, will have a higher risk of experiencing a second seizure. In general, the decision to start antiseizure medication is based on the health care provider’s assessment of many factors that influence how likely it is that another seizure will occur in that person.

Types of epilepsy

Just as there are many different kinds of seizures, there are many different kinds of epilepsy. Hundreds of different epilepsy syndromes—disorders that include seizures as a prominent symptom—have been identified. Some of these syndromes appear to be either hereditary or caused by de novo gene changes. For other syndromes, the cause is unknown. Epilepsy syndromes are frequently described by their symptoms or by where in the brain they originate.

• Absence epilepsy  is characterized by repeated seizures that cause momentary lapses of consciousness. The seizures almost always begin in childhood or adolescence and tend to run in families, suggesting that they may at least be partially due to genetic factors. Individuals may show purposeless movements during their seizures, such as a jerking arm or rapidly blinking eyes, while others may have no noticeable symptoms except for brief times when they appear to be staring off into space. Immediately after a seizure, the person can resume whatever he or she was doing. However, these seizures may occur so frequently (in some cases up to 100 or more a day) that the person cannot concentrate in school or other situations.
• Frontal lobe epilepsy  is a common epilepsy syndrome that features brief focal seizures that may occur in clusters. It can affect the part of the brain that controls movement and its seizures can cause muscle weakness or unusual, uncontrolled movement such as twisting, waving the arms or legs, eyes drifting to one side, or grimacing, and are usually associated with some loss of awareness. Seizures usually occur during sleep but may also occur while awake. 
• Temporal lobe epilepsy (TLE)  is the most common epilepsy syndrome in people who get focal seizures. These seizures are often associated with auras of nausea, emotions (such as déjà vu or fear), or unusual smell or taste. The seizure itself is a brief period of impaired consciousness which may appear as a staring spell, dream-like state, or repeated automatisms. TLE often begins in childhood or teenage years. Research has shown that repeated temporal lobe seizures are often associated with shrinkage and scarring (sclerosis) of the hippocampus. The hippocampus is a brain region that is important for memory and learning.
• Neocortical epilepsy  is characterized by seizures that originate from the cerebral cortex, or outer layer. The seizures can be either focal or generalized. Symptoms may include unusual sensations, visual hallucinations, emotional changes, muscle contractions, convulsions, and a variety of other symptoms, depending on where in the brain the seizures originate.

Types of childhood epilepsy

There are many other types of epilepsy that begin in infancy or childhood. Some childhood epilepsy syndromes tend to go into remission or stop entirely during adolescence. Other syndromes, such as juvenile myoclonic epilepsy (which features jerk-like motions upon waking) and Lennox-Gastaut syndrome are usually present for life.

For example:

• Infantile spasms are clusters of seizures that usually begin before the age of 6 months. During these seizures the infant may drop their head, bend at the waist, jerk their arms up toward their head, and/or cry out.
• Childhood absence epilepsy usually stops when the child reaches puberty. However, some children will continue to have absence seizures into adulthood and/or go on to develop other seizure types.
• Children with Lennox-Gastaut syndrome have several different types of seizures, including atonic seizures, which cause sudden falls and are also called drop attacks. Children with this condition usually begin having seizures before age four. This severe form of epilepsy can be very difficult to treat.
• Rasmussen's encephalitis  is a progressive form of epilepsy in which half the brain shows chronic inflammation.
• Children with Dravet syndrome and Tuberous Sclerosis Complex typically have seizures that start before age one.
• Hypothalamic hamartoma  is a rare form of epilepsy that first occurs during childhood and is associated with malformations of the hypothalamus at the base of the brain. People with this disorder have seizures that resemble laughing or crying. Such seizures frequently go unrecognized and are difficult to diagnose.
• Developmental and Epileptic Encephalopathy (DEE) refers to a group of severe epilepsies that are characterized both by seizures, which are often drug-resistant, as well as encephalopathy, which is a term used to describe significant developmental delay or even loss of developmental skills.

Nonepileptic seizures

Non-epileptic seizures outwardly resemble epileptic seizures but are not associated with electrical discharge in the brain. Non-epileptic events may be referred to as psychogenic non-epileptic seizures (PNES). PNES do not respond to antiseizure drugs; instead, they are often treated by cognitive behavioral therapy to decrease stress and improve mindfulness.

A history of traumatic events is among the known risk factors for PNES. People with PNES should be evaluated for underlying psychiatric disorders and treated appropriately. Some people with epilepsy have psychogenic seizures in addition to their epileptic seizures.

Other nonepileptic events may be caused by:

• Tourette syndrome
• Cardiac arrhythmia (irregular heartbeat)
• Other medical conditions with symptoms that resemble seizures

Because symptoms of these disorders can look very much like epileptic seizures, they are often mistaken for epilepsy.

How are epilepsy and seizures diagnosed and treated?

Diagnosing epilepsy and seizures.

Accurate diagnosis of epilepsy is crucial for finding an effective treatment. Several tests are used to determine whether a person has epilepsy and, if so, what kind of seizures the person has. Generally, epilepsy is diagnosed after a person has had two or more unprovoked seizures separated by at least 24 hours.

Medical history

Taking a detailed medical history, including symptoms and duration of the seizures, is still one of the best methods available to determine what kind of seizures a person has had and to help determine what type of epilepsy the person has. The medical history should include details about any past illnesses or other symptoms a person may have had, as well as any family history of seizures.

Since people who have a seizure often do not remember what happened, accounts from people who have witnessed the seizures are very important. The person who experienced the seizure is asked about whether they felt anything unique (warning experiences) before the seizure started. The observers will be asked to provide a detailed description and timeline for the seizure.

Imaging and monitoring epilepsy

There are several scans and imaging techniques that can help diagnose and monitor a person's epilepsy. These include:

• An electroencephalogram (EEG) , a test that measures electrical activity in the brain, can look for abnormalities in the person's brain waves and may help to determine if antiseizure drugs would help. Video monitoring may be used in conjunction with EEG to determine the nature of a person's seizures and to rule out other disorders that may look like epilepsy.
• SEEG (stereoelectoencephalograpy) is the surgical implantation of electrodes into the brain in order to better find where the seizures are located. SEEG can help determine if an individual is a candidate for epilepsy surgery. A magnetoencephalogram (MEG) measures the magnetic signals generated by neurons to help find unusual brain activity. MEG can help surgeons plan any appropriate surgeries to remove focal areas involved in seizures while minimizing interference with normal brain function.  
• CT (computerized tomography) and MRI (magnetic resonance imaging) scans reveal structural abnormalities of the brain such as tumors and cysts, which may cause seizures. A type of MRI called functional MRI (fMRI) can be used to localize normal brain activity and detect abnormalities in brain function.
• PET (positron emission tomography) scans take pictures of the brain and show regions of the brain with normal and abnormal chemical activity. PET scans can be used to identify brain regions with lower-than-normal metabolism, which can indicate the focus of the seizure after it has stopped.
• Single photon emission computed tomography (SPECT) is sometimes used to find the location of focal seizures in the brain. In a person admitted to the hospital for epilepsy monitoring, the SPECT blood flow tracer is injected within 30 seconds of a seizure. The images of brain blood flow at the time of the seizure are compared with blood flow images taken in between seizures. The seizure onset area shows a high blood flow region on the scan.

Blood tests

Blood tests can screen for metabolic or genetic disorders that may contribute to the seizures. They also may be used to check for underlying health conditions such as infections, lead poisoning, anemia, and diabetes that may be causing or triggering the seizures.

Developmental, neurological, and behavioral tests

Tests to measure motor abilities, behavior, and intellectual ability often are used to determine how epilepsy is affecting an individual. These tests also can provide clues about what kind of epilepsy the person has.

Treating epilepsy and seizures

Once epilepsy is diagnosed, it is important to begin treatment as soon as possible. There are many different ways to successfully control seizures. There are several treatment approaches that can be used, depending on the individual and the type of epilepsy.

Medications to treat seizures in epilepsy

The most common approach to treating epilepsy is to prescribe antiseizure medications. . More than 40 different antiseizure medications are available today, all with different benefits and side effects. Most seizures can be controlled with one drug. Combining medications may amplify side effects such as fatigue and dizziness, so doctors usually prescribe just one drug whenever possible. Combinations of drugs, however, are still sometimes necessary for some forms of epilepsy that do not respond to a single drug.

Which drug a person should be prescribed depends on many different factors, including:

• Seizure type
• Lifestyle and age
• Seizure frequency
• Drug side effects
• Medicines for other conditions

It may take several months to determine the best drug and dosage. If one treatment is unsuccessful, another may work better.

When starting any new antiseizure medication, a doctor will begin with a low dose and increase the dose as needed depending on how effective the drug is. Sometimes doctors monitor the level of the drug in a person’s blood to help determine when the optimal dosage has been reached. It may take time to find a dose that gives the best seizure control while minimizing side effects.

Side effects are often worse when first starting a new medicine and get better over time. Talk with your doctor about any side effects you experience while on medications and make sure they are aware of any other prescription or over-the-counter medications you are taking, including any herbs or supplements. Some antiseizure medications can affect how and whether other drugs work and can interact in harmful ways with other medications. Some may make hormonal birth control less effective in women. Some medications are harmful to the fetus, so women who plan to get pregnant should consult with their physician to be sure that they are using medications that are safe during pregnancy.

Discontinuing medication should always be done with supervision of a health care professional. It is very important to continue taking antiseizure medication for as long as it is prescribed. Discontinuing medication too early is one of the major reasons people who have been seizure-free start having new seizures and can lead to status epilepticus, which is potentially life threatening. Some people with epilepsy may be advised to discontinue their antiseizure drugs after two to three years have passed without a seizure. Others may be advised to wait for four to five years depending on the cause of the seizures.

While antiseizure medications are effective for many people with epilepsy, some do not respond to or are not able to take medications. Those individuals may be candidates for surgery, dietary changes, or devices to stop their seizures.

Diet and lifestyle changes in epilepsy

Some types of epilepsy may respond to changes in diet. A high-fat, high-protein, very low carbohydrate ketogenic diet is sometimes used to treat medication-resistant epilepsies. The diet induces a state known as ketosis, which means that the body shifts to breaking down fats instead of carbohydrates to survive. A ketogenic diet effectively reduces seizures for some people, especially children, with certain forms of epilepsy.

The ketogenic diet can be difficult to maintain since it requires that a person only eat certain foods and avoid many common foods that contain sugars and carbohydrates. Individuals using this diet to manage their seizures should be monitored to make sure they are getting enough nutrients. One side effect of a ketogenic diet is a buildup of uric acid in the blood, which can lead to kidney stones. A doctor or nutritionist can help people on this diet make sure they are getting the nutrients they need, and in the right amounts.

Sleep disorders are common among people with epilepsy and sleep deprivation is a powerful trigger of seizures. Treating sleep problems can help reduce seizures. People with epilepsy should practice good sleep hygiene: going to bed and getting up at the same time each day, reducing distractions in the bedroom, and avoiding big meals and exercise within a few hours of bedtime.

Surgery for epilepsy

Surgery is typically only considered after a person with epilepsy has unsuccessfully tried at least two medications to prevent seizures, or when doctors have found a brain lesion (an area of abnormal brain tissue) believed to be causing the seizures. When someone is found to be a good candidate, the surgery should be performed as soon as possible.

In considering a person’s candidacy for surgery to prevent seizures, doctors will review:

• Brain region involved
• Effect of the brain region on everyday function and behavior

Surgeons usually avoid operating in areas of the brain that are necessary for speech, movement, sensation, memory and thinking, or other important abilities.

While surgery can significantly reduce or even halt seizures for many people, any kind of surgery involves risk. Surgery for epilepsy does not always successfully reduce seizures and it can result in cognitive or personality changes as well as physical disability, even in people who are excellent candidates for it. Nonetheless, when medications fail, several studies have shown that surgery is much more likely to make someone seizure-free compared to attempts to use other medications. Anyone thinking about surgery for epilepsy should be assessed at an epilepsy center experienced in surgical techniques and should discuss the surgery’s risks and benefits with their healthcare team.

Even when surgery completely ends a person's seizures, it is important to continue taking antiseizure medication for some time, as prescribed by your health care provider. It is generally recommended that individuals continue medication for at least two years after a successful operation to avoid recurrence of seizures.

Surgical procedures for treating epilepsy disorders include:

• Surgery to remove a seizure focus involves removing the defined area of the brain where seizures originate. It is the most common type of surgery for epilepsy, which doctors may refer to as a lobectomy or lesionectomy and is appropriate only for focal seizures that originate in just one area of the brain.
• Multiple subpial transection  may be performed when seizures originate in part of the brain that cannot be removed. It involves making a series of cuts that are designed to prevent seizures from spreading into other parts of the brain while leaving the person's normal abilities intact.
• Corpus callosotomy or severing the network of neural connections between the right and left halves (hemispheres) of the brain, is done primarily in children with severe seizures that start in one half of the brain and spread to the other side. Corpus callosotomy can end drop attacks and other generalized seizures. However, the procedure does not stop seizures in the side of the brain where they originate, and these focal seizures may get worse after surgery.
• Hemispherectomy and hemispherotomy involve removing half of the brain's cortex, or outer layer. These procedures are used predominantly in children who have seizures that do not respond to medication because of damage that involves only half the brain, as in Rasmussen's encephalitis.
• Thermal ablation for epilepsy , also known as laser interstitial thermal therapy, directs energy to a specific, targeted brain region causing the seizures (the seizure focus). The energy, which is changed to thermal energy, destroys the brain cells causing the seizures. Laser ablation is less invasive than open brain surgery for treating epilepsy.

Some people may use neurostimulation devices to treat their epilepsy. These d evices deliver electrical stimulation to the brain to reduce seizure frequency:

• Vagus nerve stimulation involves surgically implanting a device under the skin of the chest. The device, which is attached by wire to the vagus nerve in the lower neck, delivers short bursts of electrical energy to the brain.
• Responsive stimulation uses an implanted device that analyzes brain activity patterns to detect a forthcoming seizure. Once detected, the device administers an intervention, such as electrical stimulation or a fast-acting drug to prevent the seizure from occurring.
• Deep brain stimulation involves surgically implanting an electrode connected to a pulse generator (similar to a pacemaker) to deliver electrical stimulation to specific areas in the brain to regulate electrical signals in neural circuits.

How can I or my loved one live with epilepsy and seizures?

Many people with epilepsy can do the same things as people without the disorder and have successful and productive lives. One-third or more of people with epilepsy, however, may have cognitive or neuropsychiatric symptoms that can negatively impact their quality of life. Many people with epilepsy who respond to treatment may go months or years without having a seizure.

People with treatment-resistant epilepsy may have as many as hundreds of seizures a day or they may have one seizure a year with sometimes disabling consequences. Having treatment-resistant epilepsy is associated with an increased risk of cognitive impairment, particularly if the seizures developed in early childhood. These impairments may be related to the underlying conditions associated with the epilepsy rather than to the epilepsy itself.

Special risks associated with epilepsy Although many people with epilepsy lead full, active lives, there is an increased risk of death or serious disability associated with epilepsy. There may be an increased risk of suicidal thoughts or actions related to some antiseizure medications that are also used to treat mania and bipolar disorder. Two life-threatening conditions associated with epilepsy are status epilepticus and sudden unexpected death in epilepsy (SUDEP). Status epilepticus is a potentially life-threatening condition in which a person either has an abnormally prolonged seizure (over five minutes) or does not fully regain consciousness between recurring seizures. Status epilepticus can be convulsive (where signs of a seizure are seen) or nonconvulsive (which cannot be seen and is diagnosed by an abnormal electroencephalogram or EEG). Nonconvulsive status epilepticus may look like a long episode of confusion, agitation, loss of consciousness, or even coma. Evidence has shown that five minutes is sufficient to damage neurons and that seizures are unlikely to end on their own, making it necessary to seek medical care immediately.    SUDEP (Sudden Unexplained Death in Epilepsy )   refers to  deaths in people with epilepsy that are not from injury, drowning, or other known causes . Most, but not all, cases of SUDEP happen during or right after a seizure. Current research on SUDEP points to abnormal brain activity that impacts heart and respiratory function. This may be due to variations in a person’s genes, particularly genes that cause epilepsy and also affect heart function. SUDEP can occur at any age and people with seizures that are difficult to control tend to have a higher incidence of SUDEP. People with epilepsy may be able to reduce their risk of SUDEP by carefully taking all antiseizure medication as prescribed and making sure they are receiving the best possible care for their epilepsy. Not taking the prescribed dosage of medication or not taking the most appropriate medication on a regular basis may increase the risk of SUDEP, especially in people who are taking more than one medication.

Mental health and stigmatization

Depression is common among people with epilepsy. In adults with epilepsy, depression and anxiety are the two most frequent mental health diagnoses. Depression or anxiety in people with epilepsy can be treated with counseling or most of the same medications used in people who do not have epilepsy. People with epilepsy should discuss their symptoms of mental health issues with their healthcare professionals so they can receive the appropriate treatments and care.

Children with epilepsy have a higher risk of developing depression and/or attention deficit hyperactivity disorder (ADHD) compared with their peers. Behavioral problems and/or mental health issues may precede the onset of seizures in some children. Counseling and support groups can help families cope with epilepsy in a positive manner.

Driving and recreation

Some states may not issue a driver’s license to a person with epilepsy. Individuals with epilepsy may be able to get an exception if they can demonstrate a period of being seizure-free, or if their seizures only happen during sleep.

The risk of seizures may also limit a person’s ability to participate in sports, exercise, or other recreational activities, including climbing, sailing, swimming, or working on ladders. There is some evidence that regular exercise may improve seizure control in some people, but this should be done under a doctor's supervision. The benefits of sports participation may outweigh the risks. Coaches and activity leaders can take appropriate safety precautions. Individuals with epilepsy should avoid dehydration, overexertion, and low blood sugar, as these problems can increase the risk of seizures.

Education and employment

By law, people with epilepsy in the U.S. cannot be denied employment or access to any educational, recreational, or other activity because of their epilepsy. However, significant barriers still exist for people with epilepsy in school and work. Antiseizure medications may cause side effects that interfere with concentration and memory. Children with epilepsy may need extra time to complete schoolwork, and they may need to have instructions or other information repeated for them. Some children with epilepsy will need special educational plans to help address their learning challenges. Teachers should be given instructions on what to do if a child in their classroom has a seizure, and parents should work with the school system to find reasonable ways to accommodate any special needs their child may have.

Pregnancy and parenthood

Epilepsy itself does not interfere with the ability to become pregnant and women who have epilepsy and take appropriate precautions have similar odds of having a healthy pregnancy and a healthy child to women without a chronic medical condition. With the appropriate selection of safe antiseizure medications during pregnancy, use of supplemental folic acid, and ideally, with pre-pregnancy planning, most people with epilepsy can have a healthy pregnancy with good outcomes for themselves and their developing child.

Women with epilepsy should be advised that some antiseizure medications carry an increased risk of birth defects. It is important to work with a team of providers that includes a neurologist and an obstetrician to learn about any special risks associated with epilepsy and antiseizure medications.

Children of parents with epilepsy have about five percent risk of developing the condition at some point, compared to roughly one percent in a child in the general population. However, the risk of developing epilepsy increases if a parent has a hereditary form of the disorder. Parents who are concerned that their epilepsy may be hereditary may wish to consult a genetics health care provider to determine their risk of passing on the disorder.

What are the latest updates on epilepsy and seizures?

NINDS conducts and supports research to better understand and diagnose epilepsy, develop new treatments, and ultimately, to prevent epilepsy. NINDS epilepsy research efforts include:

The Epilepsy Therapy Screening Program (previously called the Anticonvulsant Screening Program) was created in 1975 to facilitate the discovery of new antiseizure drugs and has contributed to the development of nearly a dozen approved medications.

NINDS Centers Without Walls (CWoW) for Collaborative Research in the Epilepsies are multicenter, multidisciplinary groups that address research challenges to advance prevention, diagnosis, and/or treatment of the epilepsies and related co-occurring conditions. NINDS’s epilepsy CWoW projects include:

• Epilepsy 4000 (Epi4K)  was an international effort to analyze DNA from 4,000 people with epilepsy and their relatives to identify disease-causing genes. Epi4K has now expanded into the worldwide effort called the Epi25 Collaborative with support from the National Human Genome Research Institute .
• The Center for SUDEP Research (CSR) brought together a collaboration of researchers with diverse expertise from multiple academic institutions in the United States and England to study and understand Sudden Unexpected Death in Epilepsy (SUDEP). 
• The Epilepsy Bioinformatics Study for Antiepileptogenic Therapy (EpiBioS4Rx) uses human and animal studies to investigate epilepsy that develops following traumatic brain injury (TBI) with the goal of identifying biomarkers (biological signs of disease) that could be used to predict who is most likely to develop epilepsy following TBI.
• The Channelopathy-Associated Epilepsy Research Center combines high-throughput technologies and high-content data modeling systems to investigate the functional consequences of genetic variants in channelopathy-associated epilepsy including Dravet Syndrome.
• The Epilepsy Multiplatform Variant Prediction (EpiMVP) aims to advance knowledge of genetic variants of uncertain significance in non-ion-channel epilepsy genes.

NINDS investigators on the NIH campus and NINDS-funded investigators around the country are conducting clinical studies aimed at finding better ways to safely detect, treat, or prevent epilepsy.

Several projects relevant to epilepsy are funded through the Brain Research through Advancing Innovative Neurotechnologies (BRAIN) Initiative®. These efforts aim to better understand, measure, and monitor how the brain generates neural activity and are working to develop new technologies and devices to measure brain activity, predict seizure onset, and deliver therapeutic stimulation to limit seizure activity.

Other areas of epilepsy research include:

• Advancing gene sequencing tools and technologies to identify genetic mutations that cause various forms of epilepsy.
• Understanding the underlying biology that contributes to epilepsy, including how brain cells communicate with one another and the role that various brain chemicals play in the disease.
• Developing new animal models to learn about the causes of epilepsy, ways to prevent the disease, and test promising therapies.
• Studying brain tissue obtained from epilepsy surgery or at the time of death to increase knowledge into causes of epilepsy and how they affect the brain.
• Identifying the genes and their function(s) relevant to rare epilepsy syndromes so that targeted treatments can be developed for children and adults with these causes of epilepsy. 
• Developing new therapies through preclinical studies utilizing the NINDS Translational research programs as well as the Blueprint Neurotherapeutics programs.
• Studying the impact of early in life seizures on cognitive and behavioral outcomes in children with epilepsy.
• Using artificial intelligence (AI) and machine learning (ML) to help to identify the specific area of the brain where the seizures begin.

How can I or my loved one help improve care for people with epilepsy and seizures?

Consider participating in a clinical trial so clinicians and scientists can learn more about epilepsy, seizures, and related disorders. Clinical research uses human study participants to help researchers learn more about a disorder and perhaps find better ways to safely detect, treat, or prevent disease.

All types of study participants are needed— those who are healthy or may have an illness or disease— of all different ages, sexes, races, and ethnicities to ensure that study results apply to as many people as possible, and that treatments will be safe and effective for everyone who will use them.

For information about participating in clinical research visit  NIH Clinical Research Trials and You . Learn about clinical trials currently looking for people with epilepsy and seizures at  Clinicaltrials.gov . 

There are additional ways for people with epilepsy and their families can help advance research. Pregnant people who are taking antiseizure drugs can help researchers learn how these drugs affect unborn children by participating in the The North American Antiepileptic Drug (AED) Pregnancy Registry . Registry participants are given educational materials on pre-conception planning and perinatal care and are asked to provide information about the health of their children, which is kept confidential and only used in an anonymized fashion by researchers.

People with epilepsy can help research efforts by making arrangements to donate tissue either at the time of surgery for epilepsy or at the time of death. Researchers can then use the tissue to study epilepsy and other disorders to better understand what causes seizures. For example, The NIH NeuroBioBank is an effort to coordinate a network of brain banks it supports in the U.S. where brain tissue and data are collected, evaluated, stored, and made available to researchers in a standardized way for the study of neurological, psychiatric and developmental disorders, including epilepsy. A l ist of participating NIH NeuroBioBank repositories and additional brain banks are maintained on the NIH NeuroBioBank website. Each brain bank may have different protocols for registering a potential donor. Individuals are strongly encouraged to contact the brain bank directly to learn more.

Where can I find more information about epilepsy and seizures? Information about epilepsy and seizures may be available from the following organizations and resources: American Epilepsy Society Phone: 312-883-3800  BeMedWise Program at NeedyMeds Phone: 978-281-6666 Caregiver Action Network Phone: 202-454-3970 or Caregiver Help Desk: 855-227-3640 Child Neurology Foundation Phone: 888-417-3435 CURE Epilepsy Phone: 312-255-1801 or 844-231-2873 Dravet Syndrome Foundation Phone: 203-392-1950 Epilepsy Alliance of America Phone: 800-642-0500 Epilepsy Leadership Council Phone: 312-883-3800 Epilepsy Foundation Phone: 800-332-1000 Family Caregiver Alliance Phone: 415-434-3388 or 800-445-8106 International League Against Epilepsy Phone: +860-586-7547 Lennox-Gastaut Syndrome (LGS) Foundation Phone: 718-374-3800 National Organization for Rare Disorders (NORD) Phone: 203-744-0100 or 800-999-6673 Partners Against Mortality in Epilepsy (PAME) The Charlie Foundation for Ketogenic Therapies Phone: 310-393-2347 The North American Antiepileptic Drug (AED) Pregnancy Registry Phone: 888-233-2334 Tuberous Sclerosis Complex Alliance Phone: 800-225-6872 or   301-562-9890 SLC6A1 Connect Phone: 303-907-8038

Learn about related topics

• Dravet Syndrome
• Febrile Seizures
• Lennox-Gastaut Syndrome

News Center

Flicker stimulation shines in clinical trial for epilepsy.

A scientist and her tools: Annabelle Singer has quantified her flicker technology with unprecedented precision in a new clinical trial. — Photo by Jerry Grillo

Biomedical engineer  Annabelle Singer  has spent the past decade developing a noninvasive therapy for Alzheimer’s disease that uses flickering lights and rhythmic tones to modulate brain waves. Now she has discovered that the technique, known as flicker, also could benefit patients with a host of other neurological disorders, from epilepsy to multiple sclerosis.

Previously, Singer and her collaborators demonstrated that the lights and sounds, delivered to patients through goggles and headphones, have beneficial effects. Flicker has been successful in animal studies and in  early human feasibility trials , where it was tested for safety, tolerance, and patient adherence.

Now, thanks to a clinical trial for people with epilepsy, the researchers quantified flicker’s effects with unprecedented precision. They also made an unexpected, but encouraging, discovery: The treatment reduced interictal epileptiform discharges (IEDs) in the brain.

These large, intermittent electrophysiological events are observed between seizures in people with epilepsy. They appear as sharp spikes on an EEG readout.

“What’s interesting about these IEDs is that they don’t just occur in epilepsy,” said Singer, McCamish Foundation Early Career Professor in the  Wallace H. Coulter Department of Biomedical Engineering at Georgia Tech and Emory University . “They occur in autism, multiple sclerosis, Alzheimer’s, and other neurological disorders, too.” And IEDs disrupt normal brain function, causing memory impairment.

Singer and her team published their findings recently in  Nature Communications .

The Rhythm in Our Heads

Inside the brain are elaborate symphonies of electrical activity: brain waves, or oscillations, that compose our memories, thoughts, and emotions. Singer wants to modulate those oscillations for therapeutic purposes. 

At specific frequencies of light and sound, the flicker treatment can induce gamma oscillations in mice. This helps the brain recruit microglia, cells responsible for removing beta amyloid, which is believed to play a central role in Alzheimer’s pathology. Part of the work is in recording what’s happening in the brain during treatment to verify how it’s working.

The patients in the trial were under the care of physician  Jon Willie  at the Emory University Hospital Epilepsy Monitoring Unit. (Willie, co-corresponding author of the study with Singer, is now at Washington University in St. Louis.) They were awaiting surgery to remove an area of the brain where seizures occur. Before that could happen, they had to undergo intracranial seizure monitoring — recording electrodes are placed in the brain to pinpoint the seizure onset zone and determine exactly which tissue should be removed. Then, patients and their care team wait for a seizure to happen. It can take days.

“In human studies, we’ve used noninvasive methods like functional MRI or scalp EEG, but they have real downsides in terms of resolution,” Singer said. “Working with these patients was a game changer. These are people with treatment-resistant epilepsy, which means that drugs aren’t working for them.”

Pathway to Healing

Singer’s team recruited 19 patients. Lead author of the study, Lou Blanpain, a former Ph.D. student in Singer’s lab and now a medical student at Emory, went from patient to patient with the flicker stimulation and recording equipment.

“Because these patients already had recording probes implanted for clinical reasons, we were able to record directly from the brain,” Singer said. “We’ve never been able to get recordings of this quality during flicker treatment before.”

As the researchers expected, flicker modulated the visual and auditory brain regions that respond strongly to stimuli. But it also reached deeper, into the medial temporal lobe and prefrontal cortex, brain regions crucial for memory. And across the brain, in regions Singer hadn’t fully explored before, she found IEDs were decreasing. 

“That has important implications for whether flicker is therapeutically relevant for people with Alzheimer’s, but also in general if we want to target anything beyond the primary sensory regions,” she said. “All of this points to the potential use of flicker in a lot of different contexts. Going forward, we’re definitely going to look at other conditions and other potential implications.”

Citation:  Lou T. Blanpain, Eric R. Cole, Emily Chen, James K. Park, Michael Y. Walelign, Robert E. Gross, Brian T. Cabaniss, Jon T. Willie, Annabelle C. Singer.  “Multisensory Flicker Modulates Widespread Brain Networks and Reduces Interictal Epileptiform Discharges,”   Nature Communications . 

Funding:  National Institutes of Health (R01 NS109226, RF1NS109226, RF1AG078736, R01 MH120194, P41 EB018783, MH12019), DARPA, McCamish Foundation, Packard Foundation.

Competing interests:  Annabelle Singer owns shares in Cognito Therapeutics, which aims to develop gamma stimulation-related products. These conflicts are managed by Georgia Tech’s Office of Research Integrity Assurance.

Jerry Grillo