EEG Signal Processing Research Articles

Multicenter intracranial EEG dataset for classification of graphoelements and artifactual signals

EEG signal processing is a fundamental method for neurophysiology research and clinical neurology practice. Historically the classification of EEG into physiological, pathological, or artifacts has been performed by expert visual review of the recordings. However, the size of EEG data recordings is rapidly increasing with a trend for higher channel counts, greater sampling frequency, and longer recording duration and complete reliance on visual data review is not sustainable. In this study, we publicly share annotated intracranial EEG data clips from two institutions: Mayo Clinic, MN, USA and St. Anne’s University Hospital Brno, Czech Republic. The dataset contains intracranial EEG that are labeled into three groups: physiological activity, pathological/epileptic activity, and artifactual signals. The dataset published here should support and facilitate training of generalized machine learning and digital signal processing methods for intracranial EEG and promote research reproducibility. Along with the data, we also propose a statistical method that is recommended for comparison of candidate classifier performance utilizing out-of-institution/out-of-patient testing.

Comprehensive review on brain-controlled mobile robots and robotic arms based on electroencephalography signals

There is a significant progress in the development of brain-controlled mobile robots and robotic arms in the recent years. New advances in electroencephalography (EEG) technology have led to the possibility of controlling external devices, such as robots, directly via the brain. The development of brain-controlled robotic devices has allowed people with bodily disabilities to enhance their mobility, individuality, and many types of activity. This paper provides a comprehensive review of EEG signal processing in robot control, including mobile robots and robotic arms, especially based on noninvasive brain computer interface systems. Various filtering approaches, feature extraction techniques, and machine learning algorithms for EEG classification are discussed and summarized. Finally, the conditions of the environments in which robots are used and robot types are also discussed.

EEG signal processing with separable convolutional neural network for automatic scoring of sleeping stage

Nowadays, among the Deep Learning works, there is a tendency to develop networks with millions of trainable parameters. However, this tendency has two main drawbacks: overfitting and resource consumption due to the low-quality features extracted by those networks. This paper presents a study focused on the scoring of sleeping EEG signals to measure if the increase of the pressure on the features due to a reduction of the number though different techniques results in a benefit. The work also studies the convenience of increasing the number of input signals in order to allow the network to extract better features. Additionally, it might be highlighted that the presented model achieves comparable results to the state-of-the-art with 1000 times less trainable and the presented model uses the whole dataset instead of the simplified versions in the published literature.

EEG data processing in ADHD diagnosis and neurofeedback

In this paper, EEG data processing was conducted in order to define the parameters for neurofeedback. A new survey was conducted based on a brief review of previous research. Two groups of participants were chosen: ADHD (3) and nonADHD (14). The main part of this study includes EEG signal data pre-processing and processing. We have outlined statistical features of observed EEG signals such as mean value, grand-mean value and their ratios. It can be concluded that an increase in grand-mean values of power theta-low beta ratio on Cz electrode gives confirmation of previous research. The value of alpha-delta power ratio higher than 1 on C3, Cz, P3, Pz, P4 in ADHD group is proposed as a new approach to classification. Based on these conclusions we will design a neurofeedback protocol as a continuation of this work.

Different Spectral Analysis Methods for the Theta/Beta Ratio Calculate Different Ratios But Do Not Distinguish ADHD from Controls

There has been ongoing research on the ratio of theta to beta power (Theta/Beta Ratio, TBR) as an EEG-based test in the diagnosis of ADHD. Earlier studies reported significant TBR differences between patients with ADHD and controls. However, a recent meta-analysis revealed a marked decline of effect size for the difference in TBR between ADHD and controls for studies published in the past decade. Here, we test if differences in EEG processing explain the heterogeneity of findings. We analyzed EEG data from two multi-center clinical studies. Five different EEG signal processing algorithms were applied to calculate the TBR. Differences between resulting TBRs were subsequently assessed for clinical usability in the iSPOT-A dataset. Although there were significant differences in the resulting TBRs, none distinguished between children with and without ADHD, and no consistent associations with ADHD symptoms arose. Different methods for EEG signal processing result in significantly different TBRs. However, none of the methods significantly distinguished between ADHD and healthy controls in our sample. The secular effect size decline for the TBR is most likely explained by factors other than differences in EEG signal processing, e.g. fewer hours of sleep in participants and differences in inclusion criteria for healthy controls.

Kalp Rahatsızlığı Riskinin Korelasyon Matrisi, Rastgele Ağaç ve Permütasyon Öznitelik Seçimi Yöntemleriyle Tespit Edilmesi

Surface EEG measurements that can be performed in hospitals and laboratories have reached a wearable and portable level with the development of today's technologies. Artificial intelligence-assisted brain-computer interface (BCI) systems play an important role in individuals with disabilities to process EEG signals and interact with the outside world. In particular, the research is becoming widespread to meet the basic needs of individuals in need of home care with an increasing population. In this study, it is aimed to design the BCI system that will detect the hunger and satiety status of the people on the computer platform through EEG measurements. In this context, a database was created by recording EEG signals with eyes open and eyes closed by 20 healthy participants in the first stage of the study. The noise of the EEG signal is eliminated by using a low pass, high pass, and notch filters. In the classification, using Wavelet Packet Transform (WPT) with Coiflet 1 and Daubechies 4 wavelets, 77.50% accuracy was achieved in eyes closed measurement, and 81% in eyes open measurement.

Gözler Açık ve Kapalı iken Alınan EEG Sinyallerinden Açlık Tokluk Durumunun Tespiti

Surface EEG measurements that can be performed in hospitals and laboratories have reached a wearable and portable level with the development of today's technologies. Artificial intelligence-assisted brain-computer interface (BCI) systems play an important role in individuals with disabilities to process EEG signals and interact with the outside world. In particular, the research is becoming widespread to meet the basic needs of individuals in need of home care with an increasing population. In this study, it is aimed to design the BCI system that will detect the hunger and satiety status of the people on the computer platform through EEG measurements. In this context, a database was created by recording EEG signals with eyes open and eyes closed by 20 healthy participants in the first stage of the study. The noise of the EEG signal is eliminated by using a low pass, high pass, and notch filters. In the classification, using Wavelet Packet Transform (WPT) with Coiflet 1 and Daubechies 4 wavelets, 77.50% accuracy was achieved in eyes closed measurement, and 81% in eyes open measurement.

A new evolutionary preprocessing approach for classification of mental arithmetic based EEG signals.

Brain computer interface systems decode brain activities from electroencephalogram (EEG) signals and translate the user's intentions into commands to control and/or communicate with augmentative or assistive devices without activating any muscle or peripheral nerve. In this paper, we aimed to improve the accuracy of these systems using improved EEG signal processing techniques through a novel evolutionary approach (fusion-based preprocessing method). This approach was inspired by chromosomal crossover, which is the transfer of genetic material between homologous chromosomes. In this study, the proposed fusion-based preprocessing method was applied to an open access dataset collected from 29 subjects. Then, features were extracted by the autoregressive model and classified by k-nearest neighbor classifier. We achieved classification accuracy (CA) ranging from 67.57 to 99.70% for the detection of binary mental arithmetic (MA) based EEG signals. In addition to obtaining an average CA of 88.71%, 93.10% of the subjects showed performance improvement using the fusion-based preprocessing method. Furthermore, we compared the proposed study with the common average reference (CAR) method and without applying any preprocessing method. The achieved results showed that the proposed method provided 3.91% and 2.75% better CA then the CAR and without applying any preprocessing method, respectively. The results also prove that the proposed evolutionary preprocessing approach has great potential to classify the EEG signals recorded during MA task.

Prediction of Pilot's Reaction Time Based on EEG Signals.

The main hypothesis of this work is that the time of delay in reaction to an unexpected event can be predicted on the basis of the brain activity recorded prior to that event. Such mental activity can be represented by electroencephalographic data. To test this hypothesis, we conducted a novel experiment involving 19 participants that took part in a 2-h long session of simulated aircraft flights. An EEG signal processing pipeline is proposed that consists of signal preprocessing, extracting bandpass features, and using regression to predict the reaction times. The prediction algorithms that are used in this study are the Least Absolute Shrinkage Operator and its Least Angle Regression modification, as well as Kernel Ridge and Radial Basis Support Vector Machine regression. The average Mean Absolute Error obtained across the 19 subjects was 114 ms. The present study demonstrates, for the first time, that it is possible to predict reaction times on the basis of EEG data. The presented solution can serve as a foundation for a system that can, in the future, increase the safety of air traffic.

EEG Signal Analysis Methods and Their Applications

EEG is a weak physiological electrical signal, which has important value in clinical and laboratory research. This paper mainly introduces several common methods of EEG signal processing, including power spectrum analysis, time-frequency analysis, bispectral analysis, etc, it mainly introduces their principles and applications in EEG signal processing, and provides methods and approaches for studying EEG.

Motor imagery EEG recognition based on conditional optimization empirical mode decomposition and multi-scale convolutional neural network

Electroencephalogram (EEG) signals classification plays a crucial role in brain computer interfaces (BCIs) system. However, the inherent complex properties of EEG signals make it challenging to get them analyzed and modeled. In this paper, a novel method based on conditional empirical mode decomposition (CEMD) and one-dimensional multi-scale convolutional neural network (1DMSCNN) is proposed to recognize motor imagery (MI) EEG signals. In the CEMD algorithm, the correlation coefficient between the original EEG signal and each intrinsic modal component (IMF) is used as the first condition to select IMFs, and the relative energy occupancy rates between the IMFs are the second condition. The CEMD algorithm is applied to remove the noise of EEG signals. Then, an EEG signals combination method is proposed to encode event-related synchronization/de-synchronization (ERS/ERD) information between the channels. Finally, a model called 1DMSCNN is built to classify the processed EEG signals. The proposed method is applied to the dataset collected in our laboratory and BCI competition IV dataset 2b. The results indicate that the proposed method can achieve higher accuracy for EEG signals classification, compared with other state-of-the-art works. In addition, the proposed algorithm is applied to the online recognition of EEG signals, a BCI system that directly interacts with brain and wheelchair is designed and implemented. This system can directly command wheelchair to turn left and right through EEG signals. The online experimental results indicate that the designed intelligent wheelchair system is a feasible BCI application. It verifies the proposed algorithm can be used in expert and intelligent systems. Our method can provide a stimulus to the development of human-robot interaction.

Current problems and development trends of specific complex methods for the diagnosis of neurotic disorders

The article is devoted to the development and testing of methods for screening and diagnosis of risk factors for the development of neurotic disorders and their initial clinical manifestations. Modern approaches to medicine are aimed at the prevention and early diagnosis of borderline mental disorders in order to provide more effective and personalized care. 87 people were examined, which, on the basis of a preliminary examination, were divided into individuals with low and high levels of neuropsychics resistance. The results of the study show significant differences in the indicators of ECG, EEG, speech signal on the psycho-activating effect between these groups. The data obtained can be used to develop a neuro-interface that has in its structure the ability to record and process ECG, EEG and speech signals, which will allow for holter monitoring of the state of mental health, thereby increasing the percentage of verified diagnosis of diseases of the neurotic spectrum.

Editorial commentary on special issue of Advances in EEG Signal Processing and Machine Learning for Epileptic Seizure Detection and Prediction.

This special issue of The Journal of Biomedical Research features novel studies on epileptic seizure detection and prediction based on advanced EEG signal processing and machine learning algorithms. The articles selected present important findings including new experimental results and theoretical studies.

A Dynamic Filtering DF-RNN Deep-Learning-Based Approach for EEG-Based Neurological Disorders Diagnosis

Filtering of unwanted signals has a great impact on the performance of EEG signal processing applied to neurological disorders diagnosis. It is so difficult to remove undesirable noises using static filtering approaches as the performance of such techniques is strongly relying on specific EEG signal sub-bands, whose locations differ from one subject to another. In this paper, we present a novel dynamic filtering approach, which makes use of Finite and Infinite Impulse Response (FIR and IIR) filters along with a Recurrent Neural Networks using a Gated-Recurrent Unit (RNN-GRU), to identify and preprocess the most informative sub-bands pertaining to a particular neurological disorder. This combination of RNN with GRU requires more hidden layers than for conventional NN structures, and therefore offers much higher capacity to learn fitting and extract features from highly complex EEG data recording to afford better harmonization of the diagnosis process. Followed by an Independent Component Analysis (ICA) algorithm, all extracted features become independent to facilitate classification of clinical disorders using Convolutional Neural Network (CNN). The proposed diagnosis system achieves an average of 100% classification accuracy for epilepsy according to an offline diagnosis process using Bonn and MIT datasets, and when the same system is applied to autism provides an average accuracy of 99.5% using KAU dataset. The presented dynamic deep-learning approach applied to EEG classification pipeline, which includes artifact removal, feature extraction and classification, leads to significant improvements in the accuracy of the diagnosis classification regarding the targeted neurological pathologies.

Python (Deep Learning and Machine Learning) for EEG Signal Processing on the Example of Recognizing the Disease of Alcoholism

Alcoholism is one of the most common diseases in the world. This type of substance abuse leads to mental and physical dependence on ethanol-containing drinks. Alcoholism is accompanied by progressive degradation of the personality and damage to the internal organs. Today still not exists a quick diagnosis method to detect this disease. This article presents the method for the quick and anonymous alcoholism diagnosis by neural networks. For this method, don't need any private information about the subject. For the implementation, we considered various algorithms of machine learning and deep neural networks. In detail analyzed the correlation of the signals from electrodes by neural networks. The wavelet transforms and the fast Fourier transform was considered. The manuscript demonstrates that the deep neural network which operates only with a dataset of EEG correlation signals can anonymously classify the alcoholic and control groups with high accuracy. On the one hand, this method will allow subjects to be tested for alcoholism without any personal data, which will not cause inconvenience or shame in the subject, and on the other hand, the subject will not be able to deceive specialists who diagnose the subject for the presence of the disease.

Review of Neural Networks in the EEG Signal Recognition

In the last decade, unprecedented progress in the development of neural networks influenced dozens of different industries, among which are signal processing for the electroencephalography process (EEG). Electroencephalography, even though it appeared in the first half of the 20th century, to this day didn’t change the physical principles of operation. But the signal processing technique due to the use of neural networks progressed significantly in this area. Evidence for this can serve that for the past 5 years more than 1000 publications on the topic of using machine learning have been published in popular libraries. Many different models of neural networks complicate the process of understanding the real situation in this area. In this manuscript, we provided the most comprehensive overview of research where were used neural networks for EEG signal processing.

Effective Hybrid Method for the Detection and Rejection of Electrooculogram (EOG) and Power Line Noise Artefacts From Electroencephalogram (EEG) Mixtures

Electrooculogram (EOG) and power line noise artefact detection and rejection have commonly utilized Stone's blind source separation (Stone's BSS) algorithm. The paper suggests a hybrid method between particle swarm optimization (PSO) and Stone's BSS for the detection and rejection of electrooculogram (EOG) and power line noise in the single-channel without the use of a notch filter. The proposed method contains three major steps: centralizing and whitening of the input EEG signal, incorporating the processing EEG signal into the iterative algorithm of the particle swarm optimization (PSO) to randomly generate the optimal value of (h <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">short</sub> , h <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">long</sub> ) and weight vector W parameters, and applying Stone's BSS using the generalized eigenvalue decomposition (GEVD) method to eliminate electrooculogram (EOG) and power line noise artefacts to obtain a clean EEG signal. The authors assess the robustness of the suggested method evaluated using real and simulated electroencephalogram (EEG) data sets. The simulated electroencephalogram (EEG) data and electrooculogram (EOG) and line noise (LN) artefacts are produced and mixed randomly in the MATLAB program; two types of real EEG data are taken in 9 and 19 channels. Evaluation results show the proposed algorithms as effective techniques for extracting both the power line noise and electrooculogram (EOG) artefacts from brain mixtures compared to specific BSS algorithms (e.g., Stone's BSS, evolutionary fast independent component analysis (EFICA), fast independent component analysis (FastICA), and joint approximate diagonalization of Eigen matrices (JADE)) while preserving the clinical features of the reconstructed EEG signal.

The Spatio-Temporal Equalization for Evoked or Event-Related Potential Detection in Multichannel EEG Data.

Evoked or Event-Related Potential (EP/ERP) detection is a major area of interest within the domain of EEG (electroencephalography) signal processing. While traditional methods of EEG processing have mostly focused on enhancing signal components, few have explored background noise suppression techniques. Optimizing the suppression of background noise can play a critical role in improving the performance of EP/ERP detection. In this study, a spatio-temporal equalization (STE) method was proposed based on the Multivariate Autoregressive (MVAR) model, which has been shown to suppress the spatio-temporal correlation of background noise and improve the EEG signal detection performance. For practical applications, two optimization schemes based on the spatio-temporal equalization method were designed to solve two common challenges in EEG signal detection: P300 and steady state visual evoked potentials. Our results demonstrated that the STE method effectively improves recognition performance of evoked or event-related potential detection. Additionally, the STE method offers fewer parameters, lower computational complexity, and easier implementation. These attributes allow the STE approach to be extended as a preprocessing method which can be used in combination with existing techniques.

Prediction of working memory ability based on EEG by functional data analysis

BackgroundThere is always a demand for fast and accurate algorithms for EEG signal processing. Owing to the high sample rate, EEG signals usually come with a large number of sample points, making it difficult to predict the working memory ability in cognitive research with EEG. New MethodFollowing well-designed experiments, the functional linear model provides a simple framework for regressions involving EEG signal predictors. The use of a data-driven basis in a linear structure naturally extends the standard linear regression model. The proposed approach utilizes B-spline approximation of functional principal components that greatly facilitates implementation. ResultsUsing LASSO feature selection, critical features have been extracted from eight frontal electrodes, and the R-square of 0.72 indicates rather strong linear association of actual observations and out-of-sample predictions. Comparison with Existing MethodsThere does not seem to be any existing methods of predicting working memory ability from N-back task tests via EEG signals; the data-driven functional linear regression method proposed in this work is, to the best of our knowledge, the first of its kind. ConclusionsThe data analytics suggest that a multiple functional linear regression model for the predictive relationship between working memory ability and frontal activity of the brain is both feasible and accurate via EEG signal processing.

A Noise-Aware Real-Time Processing Approach for Electroencephalogram Signal Classification

Electroencephalogram (EEG) signal processing has emerged as a critical problem for biometric applications due to its real-time requirement. While compressive sensing is an efficient method for signal compression, its application in EEG signal processing is limited due to its noise unawareness during transmission and time-consuming reconstruction procedure. In this paper, we propose a noise-aware sparse Bayesian learning approach with block structure (NA-BSBL) to achieve higher efficiency on data compression, reconstruction and classification. By applying novel structure for parameter and introducing the Mahalanobis Distance, our approach achieves an almost 20% reconstruction performance lift and 10% accuracy lift under noise condition. For further application of reconstructed EEG signal, we extract both the spatial and frequency domain features for classification. Experimental results show that the proposed approach can achieve 94% classification accuracy with 16% speed up compared with the conventional approach.