Log Energy Entropy-Based EEG Classification with Multilayer Neural Networks in Seizure

EEG-based fatigue state evaluation by combining complex network and frequency-spatial features

BACKGROUND

The proportion of traffic accidents caused by fatigue driving is increasing year by year, which has aroused wide concerns for researchers. In order to rapidly and accurately detect drivers' fatigue, this paper proposed an electroencephalogram (EEG)-based fatigue state evaluation method by combining complex network and frequency-spatial features.

NEW METHOD

First, this paper constructed a complex network model based on the relative wavelet entropy to characterize the correlation strength information between channels. Then, the differential entropy and symmetry quotient were respectively calculated to extract frequency and spatial features. Then, the brain heat map combined the complex network and frequency-spatial features with different dimensions together as the fusion features. Finally, a convolutional neural network-long short-term memory (CNN-LSTM) neural network was used to evaluate the three-class fatigue states of the EEG data in the Shanghai Jiao Tong University (SJTU) Emotion EEG Dataset (SEED)-VIG dataset, and it was validated on the dataset on the Mendeley Data website.

RESULTS

The experimental results of SEED-VIG dataset show that the average classification accuracy of three-class fatigue states, namely, awake, tired and drowsy, reaches 96.57 %. The average classification accuracy on the dataset on the Mendeley Data website reaches 99.23 %.

COMPARISON WITH EXISTING METHODS

This method has a best evaluation performance compared with the state-of-the-art methods for the three-class fatigue states recognition.

CONCLUSIONS

The experiment results validated the feasibility of the fatigue state evaluation method based on the correlations between channels and the frequency-spatial features, which is of great significance for developing a driver fatigue detection system.

Detectability in Scalp EEGs of Epileptic Spikes Emitted from Brain Electrical Sources of Different Sizes and Locations: A Simulation Study Using Realistic Head Models of Elderly Adults

Background. Epilepsy is prevalent in the elderly, whose brain morphologies and skull electrical characteristics differ from those of younger adults. Here, using a multivariate definition of signal-to-noise ratio (SNR), we explored the detectability of epileptic spikes in scalp EEG measurements in elderly by forward simulations of hypersynchronous spikes generated at 78 cortical regions of interest (ROIs) in the presence of background noise. Methods. Simulated electric potentials were measured at 18, 35, and 70 standard 10-20 electrode positions using three reference methods: infinity reference (INF), common average reference (CAR), and average mastoid reference (M1M2). MRIs of six elderly subjects were used to construct finite element method (FEM) models with age-adjusted skull conductivities. Results. SNRs of epileptic spikes increased with increasing sizes of the brain electrical source areas, although medial and deep brain regions such as the hippocampus showed lower SNRs, consistent with clinical findings. The SNRs were greater in the 70-channel dataset than in the 18-channel and 35-channel datasets, especially for ROIs located closer to the head surface. In addition, the SNRs were lower for the CAR and M1M2 references than for the ideal INF reference. Moreover, we found comparable results in the standard FEM heads with age-adjusted skull conductivities. Conclusions. The results provide insights for evaluating scalp EEG data in elderly patients with suspected epilepsy, and suggest that age-adjusted skull conductivity is an important factor for forward models in elderly adults, and that the standard FEM head with age-adjusted skull conductivity can be used when MRIs are not available.

The Application of Entropy in Motor Imagery Paradigms of Brain-Computer Interfaces

Background: In motor imagery brain-computer interface (MI-BCI) research, electroencephalogram (EEG) signals are complex and nonlinear. This complexity and nonlinearity render signal processing and classification challenging when employing traditional linear methods. Information entropy, with its intrinsic nonlinear characteristics, effectively captures the dynamic behavior of EEG signals, thereby addressing the limitations of traditional methods in capturing linear features. However, the multitude of entropy types leads to unclear application scenarios, with a lack of systematic descriptions. Methods: This study conducted a review of 63 high-quality research articles focused on the application of entropy in MI-BCI, published between 2019 and 2023. It summarizes the names, functions, and application scopes of 13 commonly used entropy measures. Results: The findings indicate that sample entropy (16.3%), Shannon entropy (13%), fuzzy entropy (12%), permutation entropy (9.8%), and approximate entropy (7.6%) are the most frequently utilized entropy features in MI-BCI. The majority of studies employ a single entropy feature (79.7%), with dual entropy (9.4%) and triple entropy (4.7%) being the most prevalent combinations in multiple entropy applications. The incorporation of entropy features can significantly enhance pattern classification accuracy (by 8-10%). Most studies (67%) utilize public datasets for classification verification, while a minority design and conduct experiments (28%), and only 5% combine both methods. Conclusions: Future research should delve into the effects of various entropy features on specific problems to clarify their application scenarios. As research methodologies continue to evolve and advance, entropy features are poised to play a significant role in a wide array of fields and contexts.

The Electroencephalogram (EEG) Study for Estimating Endurance Sports Performance Base on Eigenvalues Extraction Method

OBJECTIVES

Brain-behavior connections are a new means to evaluate sports performance. This electroencephalogram (EEG) study aims to estimate endurance exercise performance by investigating eigenvalue trends and comparing their sensitivity and linearity.

METHODS

Twenty-three cross-country skiers completed endurance cycling tasks. Twenty-four-channel full-brain EEG signals were recorded in the motor phase and recovery phase continuously. Eighteen EEG eigenvalues calculation methods were collected, commonly used in previous research. Time-frequency, band power, and nonlinear analyses were used to calculate the EEG eigenvalues. Their regression coefficients and correlation coefficients were calculated and compared, with the linear regression method.

RESULTS

The time-frequency eigenvalues shift slightly throughout the test. The power eigenvalues changed significantly before and after motor and recovery, but the linearity was not satisfactory. The sensitivity and linearity of the nonlinear eigenvalues were stronger than the other eigenvalues. Of all eigenvalues, Shannon entropy showed completely non-overlapping distribution intervals in the regression coefficients of the two phases, which were -0.1474 ± 0.0806 s-1 in the motor phase and 0.2560 ± 0.1365 s-1 in the recovery phase. Shannon entropy amplitude decreased more in the F region of the brain than in the other regions. Additionally, the higher the level of sport, the slower the decline in Shannon entropy of the athlete.

CONCLUSIONS

The Shannon entropy method provided more accurate estimations for endurance exercise performance compared to other eigenvalues.

The EEG-Based Fusion Entropy-Featured Identification of Isometric Contraction Forces under the Same Action

This study explores the important role of assessing force levels in accurately controlling upper limb movements in human-computer interfaces. It uses a new method that combines entropy to improve the recognition of force levels. This research aims to differentiate between different levels of isometric contraction forces using electroencephalogram (EEG) signal analysis. It integrates eight different entropy measures: power spectrum entropy (PSE), singular spectrum entropy (SSE), logarithmic energy entropy (LEE), approximation entropy (AE), sample entropy (SE), fuzzy entropy (FE), alignment entropy (PE), and envelope entropy (EE). The findings emphasize two important advances: first, including a wide range of entropy features significantly improves classification efficiency; second, the fusion entropy method shows exceptional accuracy in classifying isometric contraction forces. It achieves an accuracy rate of 91.73% in distinguishing between 15% and 60% maximum voluntary contraction (MVC) forces, along with 69.59% accuracy in identifying variations across 15%, 30%, 45%, and 60% MVC. These results illuminate the efficacy of employing fusion entropy in EEG signal analysis for isometric contraction detection, heralding new opportunities for advancing motor control and facilitating fine motor movements through sophisticated human-computer interface technologies.

Classification of First-Episode Psychosis with EEG Signals: ciSSA and Machine Learning Approach

First-episode psychosis (FEP) typically marks the onset of severe psychiatric disorders and represents a critical period in the field of mental health. The early diagnosis of this condition is essential for timely intervention and improved clinical outcomes. In this study, the classification of FEP was investigated using the analysis of electroencephalography (EEG) signals and circulant spectrum analysis (ciSSA) sub-band signals. FEP poses a significant diagnostic challenge in the realm of mental health, and it is aimed at introducing a novel and effective approach for early diagnosis. To achieve this, the LASSO method was utilized to select the most significant features derived from entropy, frequency, and statistical-based characteristics obtained from ciSSA sub-band signals, as well as their hybrid combinations. Subsequently, a high-performance classification model has been developed using machine learning techniques, including ensemble, support vector machine (SVM), and artificial neural network (ANN) methods. The results of this study demonstrated that the hybrid features extracted from EEG signals' ciSSA sub-bands, in combination with the SVM method, achieved a high level of performance, with an area under curve (AUC) of 0.9893, an accuracy of 96.23%, a sensitivity of 0.966, a specificity of 0.956, a precision of 0.9667, and an F1 score of 0.9666. This has revealed the effectiveness of the ciSSA-based method for classifying FEP from EEG signals.

Using machine learning methods and EEG to discriminate aircraft pilot cognitive workload during flight

Pilots of aircraft face varying degrees of cognitive workload even during normal flight operations. Periods of low cognitive workload may be followed by periods of high cognitive workload and vice versa. During such changing demands, there exists potential for increased error on behalf of the pilots due to periods of boredom or excessive cognitive task demand. To further understand cognitive workload in aviation, the present study involved collection of electroencephalogram (EEG) data from ten (10) collegiate aviation students in a live-flight environment in a single-engine aircraft. Each pilot possessed a Federal Aviation Administration (FAA) commercial pilot certificate and either FAA class I or class II medical certificate. Each pilot flew a standardized flight profile representing an average instrument flight training sequence. For data analysis, we used four main sub-bands of the recorded EEG signals: delta, theta, alpha, and beta. Power spectral density (PSD) and log energy entropy of each sub-band across 20 electrodes were computed and subjected to two feature selection algorithms (recursive feature elimination (RFE) and lasso cross-validation (LassoCV), and a stacking ensemble machine learning algorithm composed of support vector machine, random forest, and logistic regression. Also, hyperparameter optimization and tenfold cross-validation were used to improve the model performance, reliability, and generalization. The feature selection step resulted in 15 features that can be considered an indicator of pilots' cognitive workload states. Then these features were applied to the stacking ensemble algorithm, and the highest results were achieved using the selected features by the RFE algorithm with an accuracy of 91.67% (± 0.11), a precision of 93.89% (± 0.09), recall of 91.67% (± 0.11), F-score of 91.22% (± 0.12), and the mean ROC-AUC of 0.93 (± 0.06). The achieved results indicated that the combination of PSD and log energy entropy, along with well-designed machine learning algorithms, suggest the potential for the use of EEG to discriminate periods of the low, medium, and high workload to augment aircraft system design, including flight automation features to improve aviation safety.

Wavelet-Based Classification of Enhanced Melanoma Skin Lesions through Deep Neural Architectures

In recent years, skin cancer diagnosis has been aided by the most sophisticated and advanced machine learning algorithms, primarily implemented in the spatial domain. In this research work, we concentrated on two crucial phases of a computer-aided diagnosis system: (i) image enhancement through enhanced median filtering algorithms based on the range method, fuzzy relational method, and similarity coefficient, and (ii) wavelet decomposition using DB4, Symlet, RBIO, and extracting seven unique entropy features and eight statistical features from the segmented image. The extracted features were then normalized and provided for classification based on supervised and deep-learning algorithms. The proposed system is comprised of enhanced filtering algorithms, Normalized Otsu’s Segmentation, and wavelet-based entropy. Statistical feature extraction led to a classification accuracy of 93.6%, 0.71% higher than the spatial domain-based classification. With better classification accuracy, the proposed system will assist clinicians and dermatology specialists in identifying skin cancer early in its stages.

An Intelligent Motor Imagery Detection System Using Electroencephalography with Adaptive Wavelets

Classification of motor imagery (MI) tasks provides a robust solution for specially-abled people to connect with the milieu for brain-computer interface. Precise selection of uniform tuning parameters of tunable Q wavelet transform (TQWT) for electroencephalography (EEG) signals is arduous. Therefore, this paper proposes robust TQWT for automatically selecting optimum tuning parameters to decompose non-stationary EEG signals accurately. Three evolutionary optimization algorithms are explored for automating the tuning parameters of robust TQWT. The fitness function of the mean square error of decomposition is used. This paper also exploits channel selection using a Laplacian score for dominant channel selection. Important features elicited from sub-bands of robust TQWT are classified using different kernels of the least square support vector machine classifier. The radial basis function kernel has provided the highest accuracy of 99.78%, proving that the proposed method is superior to other state-of-the-art using the same database.

Automated Emotion Identification Using Fourier-Bessel Domain-Based Entropies

Human dependence on computers is increasing day by day; thus, human interaction with computers must be more dynamic and contextual rather than static or generalized. The development of such devices requires knowledge of the emotional state of the user interacting with it; for this purpose, an emotion recognition system is required. Physiological signals, specifically, electrocardiogram (ECG) and electroencephalogram (EEG), were studied here for the purpose of emotion recognition. This paper proposes novel entropy-based features in the Fourier-Bessel domain instead of the Fourier domain, where frequency resolution is twice that of the latter. Further, to represent such non-stationary signals, the Fourier-Bessel series expansion (FBSE) is used, which has non-stationary basis functions, making it more suitable than the Fourier representation. EEG and ECG signals are decomposed into narrow-band modes using FBSE-based empirical wavelet transform (FBSE-EWT). The proposed entropies of each mode are computed to form the feature vector, which are further used to develop machine learning models. The proposed emotion detection algorithm is evaluated using publicly available DREAMER dataset. K-nearest neighbors (KNN) classifier provides accuracies of 97.84%, 97.91%, and 97.86% for arousal, valence, and dominance classes, respectively. Finally, this paper concludes that the obtained entropy features are suitable for emotion recognition from given physiological signals.

A Review of Recent Developments in Driver Drowsiness Detection Systems

Continuous advancements in computing technology and artificial intelligence in the past decade have led to improvements in driver monitoring systems. Numerous experimental studies have collected real driver drowsiness data and applied various artificial intelligence algorithms and feature combinations with the goal of significantly enhancing the performance of these systems in real-time. This paper presents an up-to-date review of the driver drowsiness detection systems implemented over the last decade. The paper illustrates and reviews recent systems using different measures to track and detect drowsiness. Each system falls under one of four possible categories, based on the information used. Each system presented in this paper is associated with a detailed description of the features, classification algorithms, and used datasets. In addition, an evaluation of these systems is presented, in terms of the final classification accuracy, sensitivity, and precision. Furthermore, the paper highlights the recent challenges in the area of driver drowsiness detection, discusses the practicality and reliability of each of the four system types, and presents some of the future trends in the field.

COVID-19 detection in cough, breath and speech using deep transfer learning and bottleneck features

We present an experimental investigation into the effectiveness of transfer learning and bottleneck feature extraction in detecting COVID-19 from audio recordings of cough, breath and speech. This type of screening is non-contact, does not require specialist medical expertise or laboratory facilities and can be deployed on inexpensive consumer hardware such as a smartphone. We use datasets that contain cough, sneeze, speech and other noises, but do not contain COVID-19 labels, to pre-train three deep neural networks: a CNN, an LSTM and a Resnet50. These pre-trained networks are subsequently either fine-tuned using smaller datasets of coughing with COVID-19 labels in the process of transfer learning, or are used as bottleneck feature extractors. Results show that a Resnet50 classifier trained by this transfer learning process delivers optimal or near-optimal performance across all datasets achieving areas under the receiver operating characteristic (ROC AUC) of 0.98, 0.94 and 0.92 respectively for all three sound classes: coughs, breaths and speech. This indicates that coughs carry the strongest COVID-19 signature, followed by breath and speech. Our results also show that applying transfer learning and extracting bottleneck features using the larger datasets without COVID-19 labels led not only to improved performance, but also to a marked reduction in the standard deviation of the classifier AUCs measured over the outer folds during nested cross-validation, indicating better generalisation. We conclude that deep transfer learning and bottleneck feature extraction can improve COVID-19 cough, breath and speech audio classification, yielding automatic COVID-19 detection with a better and more consistent overall performance.

Detection of Movement Intention in EEG-Based Brain-Computer Interfaces Using Fourier-Based Synchrosqueezing Transform

Patients with motor impairments need caregivers' help to initiate the operation of brain-computer interfaces (BCI). This study aims to identify and characterize movement intention using multichannel electroencephalography (EEG) signals as a means to initiate BCI systems without extra accessories/methodologies. We propose to discriminate the resting and motor imagery (MI) states with high accuracy using Fourier-based synchrosqueezing transform (FSST) as a feature extractor. FSST has been investigated and compared with other popular approaches in 28 healthy subjects for a total of 6657 trials. The accuracy and f-measure values were obtained as 99.8% and 0.99, respectively, when FSST was used as the feature extractor and singular value decomposition (SVD) as the feature selection method and support vector machines as the classifier. Moreover, this study investigated the use of data that contain certain amount of noise without any preprocessing in addition to the clean counterparts. Furthermore, the statistical analysis of EEG channels with the best discrimination (of resting and MI states) characteristics demonstrated that F4-Fz-C3-Cz-C4-Pz channels and several statistical features had statistical significance levels, [Formula: see text], less than 0.05. This study showed that the preparation of the movement can be detected in real-time employing FSST-SVD combination and several channels with minimal pre-processing effort.

A self-learned decomposition and classification model for schizophrenia diagnosis

BACKGROUND

Schizophrenia (SZ) is a type of neurological disorder that is diagnosed by professional psychiatrists based on interviews and manual screening of patients. The procedures are time-consuming, burdensome, and prone to human error. This urgently necessitates the development of an effective and precise computer-aided design for the detection of SZ. One such efficient source for SZ detection is the electroencephalogram (EEG) signals. Because EEG signals are non-stationary, it is challenging to find representative information in its raw form. Decomposing the signals into multi-modes can provide detailed insight information from it. But the choice of uniform decomposition and hyper-parameters leads to information loss affecting system performance drastically.

METHOD

In this paper, automatic signal decomposition and classification methods are proposed for the detection of SZ and healthy control EEG signals. The Fisher score method is used for the selection of the most discriminant channel. Flexible tunable Q wavelet transform (F-TQWT) is developed for efficient decomposition of EEG signals by reducing root mean square error with grey wolf optimization (GWO) algorithm. Five features are extracted from the adaptively generated subbands and selected by the Kruskal Wallis test. The feature matrix is given as an input to the flexible least square support vector machine (F-LSSVM) classifier. The hyper-parameters and kernel of classifier are selected such that the accuracy of each subband is maximized using GWO algorithm.

RESULTS

The effectiveness and superiority of the proposed method is tested by evaluating seven performance parameters. An accuracy of 91.39%, sensitivity, specificity, precision, F-1 measure, false positive rate and error of 92.65%, 93.22%, 95.57%, 0.9306, 6.78% and 8.61% is achieved. The results prove superiority of the developed F-TQWT decomposition and F-LSSVM classifier over existing methodologies.

CONCLUSION

The EEG signals of healthy control and SZ subjects performing motor and auditory tasks simultaneously provide higher discrimination ability over the subjects performing auditory and motory tasks separately. The developed model is accurate, robust, and effective as it is developed on a relatively larger data-set, obtained maximum performance, and tested using ten-fold cross-validation technique. This proposed model is ready to be put to test for real-time SZ detection.

Personalized Evaluation of Atrial Complexity of Patients Undergoing Atrial Fibrillation Ablation: A Clinical Computational Study

Current clinical guidelines establish Pulmonary Vein (PV) isolation as the indicated treatment for Atrial Fibrillation (AF). However, AF can also be triggered or sustained due to atrial drivers located elsewhere in the atria. We designed a new simulation workflow based on personalized computer simulations to characterize AF complexity of patients undergoing PV ablation, validated with non-invasive electrocardiographic imaging and evaluated at one year after ablation. We included 30 patients using atrial anatomies segmented from MRI and simulated an automata model for the electrical modelling, consisting of three states (resting, excited and refractory). In total, 100 different scenarios were simulated per anatomy varying rotor number and location. The 3 states were calibrated with Koivumaki action potential, entropy maps were obtained from the electrograms and compared with ECGi for each patient to analyze PV isolation outcome. The completion of the workflow indicated that successful AF ablation occurred in patients with rotors mainly located at the PV antrum, while unsuccessful procedures presented greater number of driving sites outside the PV area. The number of rotors attached to the PV was significantly higher in patients with favorable long-term ablation outcome (1-year freedom from AF: 1.61 ± 0.21 vs. AF recurrence: 1.40 ± 0.20; p-value = 0.018). The presented workflow could improve patient stratification for PV ablation by screening the complexity of the atria.

COVID-19 cough classification using machine learning and global smartphone recordings

We present a machine learning based COVID-19 cough classifier which can discriminate COVID-19 positive coughs from both COVID-19 negative and healthy coughs recorded on a smartphone. This type of screening is non-contact, easy to apply, and can reduce the workload in testing centres as well as limit transmission by recommending early self-isolation to those who have a cough suggestive of COVID-19. The datasets used in this study include subjects from all six continents and contain both forced and natural coughs, indicating that the approach is widely applicable. The publicly available Coswara dataset contains 92 COVID-19 positive and 1079 healthy subjects, while the second smaller dataset was collected mostly in South Africa and contains 18 COVID-19 positive and 26 COVID-19 negative subjects who have undergone a SARS-CoV laboratory test. Both datasets indicate that COVID-19 positive coughs are 15%–20% shorter than non-COVID coughs. Dataset skew was addressed by applying the synthetic minority oversampling technique (SMOTE). A leave-p-out cross-validation scheme was used to train and evaluate seven machine learning classifiers: logistic regression (LR), k-nearest neighbour (KNN), support vector machine (SVM), multilayer perceptron (MLP), convolutional neural network (CNN), long short-term memory (LSTM) and a residual-based neural network architecture (Resnet50). Our results show that although all classifiers were able to identify COVID-19 coughs, the best performance was exhibited by the Resnet50 classifier, which was best able to discriminate between the COVID-19 positive and the healthy coughs with an area under the ROC curve (AUC) of 0.98. An LSTM classifier was best able to discriminate between the COVID-19 positive and COVID-19 negative coughs, with an AUC of 0.94 after selecting the best 13 features from a sequential forward selection (SFS). Since this type of cough audio classification is cost-effective and easy to deploy, it is potentially a useful and viable means of non-contact COVID-19 screening.

A Review of EEG Signal Features and their Application in Driver Drowsiness Detection Systems

Detecting drowsiness in drivers, especially multi-level drowsiness, is a difficult problem that is often approached using neurophysiological signals as the basis for building a reliable system. In this context, electroencephalogram (EEG) signals are the most important source of data to achieve successful detection. In this paper, we first review EEG signal features used in the literature for a variety of tasks, then we focus on reviewing the applications of EEG features and deep learning approaches in driver drowsiness detection, and finally we discuss the open challenges and opportunities in improving driver drowsiness detection based on EEG. We show that the number of studies on driver drowsiness detection systems has increased in recent years and that future systems need to consider the wide variety of EEG signal features and deep learning approaches to increase the accuracy of detection.

A new approach: information gain algorithm-based k-nearest neighbors hybrid diagnostic system for Parkinson's disease

Parkinson's disease (PD) is a slow and insidiously progressive neurological brain disorder. The development of expert systems capable of automatically and highly accurately diagnosing early stages of PD based on speech signals would provide an important contribution to the health sector. For this purpose, the Information Gain Algorithm-based K-Nearest Neighbors (IGKNN) model was developed. This approach was applied to the feature data sets formed using the Tunable Q-factor Wavelet Transform (TQWT) method. First, 12 sub-feature data sets forming the TQWT feature group were analyzed separately after which the one with the best performance was selected, and the IGKNN model was applied to this sub-feature data set. Finally, it was observed that the performance results provided with the IGKNN system for this sub-feature data set were better than those for the complete set of data. According to the results, values of receiver operating characteristic and precision-recall curves exceeded 0.95, and a classification accuracy of almost 98% was obtained with the 22 features selected from this sub-group. In addition, the kappa coefficient was 0.933 and showed a perfect agreement between actual and predicted values. The performance of the IGKNN system was also compared with results from other studies in the literature in which the same data were used, and the approach proposed in this study far outperformed any approaches reported in the literature. Also, as in this IGKNN approach, an expert system that can diagnose PD and achieve maximum performance with fewer features from the audio signals has not been previously encountered.

Approximate Entropy of Brain Network in the Study of Hemispheric Differences

Human brain, a dynamic complex system, can be studied with different approaches, including linear and nonlinear ones. One of the nonlinear approaches widely used in electroencephalographic (EEG) analyses is the entropy, the measurement of disorder in a system. The present study investigates brain networks applying approximate entropy (ApEn) measure for assessing the hemispheric EEG differences; reproducibility and stability of ApEn data across separate recording sessions were evaluated. Twenty healthy adult volunteers were submitted to eyes-closed resting EEG recordings, for 80 recordings. Significant differences in the occipital region, with higher values of entropy in the left hemisphere than in the right one, show that the hemispheres become active with different intensities according to the performed function. Besides, the present methodology proved to be reproducible and stable, when carried out on relatively brief EEG epochs but also at a 1-week distance in a group of 36 subjects. Nonlinear approaches represent an interesting probe to study the dynamics of brain networks. ApEn technique might provide more insight into the pathophysiological processes underlying age-related brain disconnection as well as for monitoring the impact of pharmacological and rehabilitation treatments.

Robust Single-Trial EEG-Based Authentication Achieved with a 2-Stage Classifier

The risk of personal data exposure through unauthorized access has never been as imminent as today. To counter this, biometric authentication has been proposed: the use of distinctive physiological and behavioral characteristics as a form of identification and access control. One of the recent developments is electroencephalography (EEG)-based authentication. It builds on the subject-specific nature of brain responses which are difficult to recreate artificially. We propose an authentication system based on EEG signals recorded in response to a simple motor paradigm. Authentication is achieved with a novel two-stage decoder. In the first stage, EEG signal features are extracted using an inception- and a VGG-like deep learning neural network (NN) both of which we compare with principal component analysis (PCA). In the second stage, a support vector machine (SVM) is used for binary classification to authenticate the subject based on the extracted features. All decoders are trained on EEG motor-movement data recorded from 105 subjects. We achieved with the VGG-like NN-SVM decoder a false-acceptance rate (FAR) of 2.55% with an overall accuracy of 88.29%, a FAR of 3.33% with an accuracy of 87.47%, and a FAR of 2.89% with an accuracy of 90.68% for 8, 16, and 64 channels, respectively. With the Inception-like NN-SVM decoder we achieved a false-acceptance rate (FAR) of 4.08% with an overall accuracy of 87.29%, a FAR of 3.53% with an accuracy of 85.31%, and a FAR of 1.27% with an accuracy of 93.40% for 8, 16, and 64 channels, respectively. The PCA-SVM decoder achieved accuracies of 92.09%, 92.36%, and 95.64% with FARs of 2.19%, 2.17%, and 1.26% for 8, 16, and 64 channels, respectively.

Detection of Focal and Non-Focal Electroencephalogram Signals Using Fast Walsh-Hadamard Transform and Artificial Neural Network

The discrimination of non-focal class (NFC) and focal class (FC), is vital in localizing the epileptogenic zone (EZ) during neurosurgery. In the conventional diagnosis method, the neurologist has to visually examine the long hour electroencephalogram (EEG) signals, which consumes time and is prone to error. Hence, in this present work, automated diagnosis of FC EEG signals from NFC EEG signals is developed using the Fast Walsh-Hadamard Transform (FWHT) method, entropies, and artificial neural network (ANN). The FWHT analyzes the EEG signals in the frequency domain and decomposes it into the Hadamard coefficients. Five different nonlinear features, namely approximate entropy (ApEn), log-energy entropy (LogEn), fuzzy entropy (FuzzyEn), sample entropy (SampEn), and permutation entropy (PermEn) are extracted from the decomposed Hadamard coefficients. The extracted features detail the nonlinearity in the NFC and the FC EEG signals. The judicious entropy features are supplied to the ANN classifier, with a 10-fold cross-validation method to classify the NFC and FC classes. Two publicly available datasets such as the University of Bonn and Bern-Barcelona dataset are used to evaluate the proposed approach. A maximum sensitivity of 99.70%, the accuracy of 99.50%, and specificity of 99.30% with the 3750 pairs of NFC and FC signal are achieved using the Bern-Barcelona dataset, while the accuracy of 92.80%, the sensitivity of 91%, and specificity of 94.60% is achieved using University of Bonn dataset. Compared to the existing technique, the proposed approach attained a maximum classification performance in both the dataset.

Identification of epilepsy from intracranial EEG signals by using different neural network models

In this work, a framework is provided for identifying intracranial electroencephalography (iEEG) seizures based on discrete wavelet transform (DWT) analysis of iEEG signals using forward propagation and feedback neural networks. The performance of 5 different data sets combination classifications is studied using the probabilistic neural network (PNN), learning vector quantization neural network (LVQ) and Elman neural network (ENN). Different feature combinations serve as the input vectors of the classifiers to obtain the best outcomes. It has been found that PNN has less running time and provides better classification accuracy (CA) than ENN and LVQ classifiers for all 5 classification problems. It is worth noticing that the CA for the C-D classification task, which shows the status of pre-ictal versus post-ictal, has been greatly improved, and reached 83.13%. Hence, the epilepsy iEEG signals pattern recognition based on DWT statistical features using the PNN classifier is more suitable for forming a reliable, automatic classification system in order to assist doctors in diagnosis.

Current Status, Challenges, and Possible Solutions of EEG-Based Brain-Computer Interface: A Comprehensive Review

Brain-Computer Interface (BCI), in essence, aims at controlling different assistive devices through the utilization of brain waves. It is worth noting that the application of BCI is not limited to medical applications, and hence, the research in this field has gained due attention. Moreover, the significant number of related publications over the past two decades further indicates the consistent improvements and breakthroughs that have been made in this particular field. Nonetheless, it is also worth mentioning that with these improvements, new challenges are constantly discovered. This article provides a comprehensive review of the state-of-the-art of a complete BCI system. First, a brief overview of electroencephalogram (EEG)-based BCI systems is given. Secondly, a considerable number of popular BCI applications are reviewed in terms of electrophysiological control signals, feature extraction, classification algorithms, and performance evaluation metrics. Finally, the challenges to the recent BCI systems are discussed, and possible solutions to mitigate the issues are recommended.

Cross-database evaluation of EEG based epileptic seizures detection driven by adaptive median feature baseline correction

OBJECTIVE

In long-term electroencephalogram (EEG) signals, automated classification of epileptic seizures is desirable in diagnosing epilepsy patients, as it otherwise depends on visual inspection. To the best of the author's knowledge, existing studies have validated their algorithms using cross-validation on the same database and less number of attempts have been made to extend their work on other databases to test the generalization capability of the developed algorithms. In this study, we present the algorithm for cross-database evaluation for classification of epileptic seizures using five EEG databases collected from different centers. The cross-database framework helps when sufficient epileptic seizures EEG data are not available to build automated seizure detection model.

METHODS

Two features, namely successive decomposition index and matrix determinant were extracted at a segmentation length of 4 s (50% overlap). Then, adaptive median feature baseline correction (AM-FBC) was applied to overcome the inter-patient and inter-database variation in the feature distribution. The classification was performed using a support vector machine classifier with leave-one-database-out cross-validation. Different classification scenarios were considered using AM-FBC, smoothing of the train and test data, and post-processing of the classifier output.

RESULTS

Simulation results revealed the highest area under the curve-sensitivity-specificity-false detections (per hour) of 1-1-1-0.15, 0.89-0.99-0.82-2.5, 0.99-0.73-1-1, 0.95-0.97-0.85-1.7, 0.99-0.99-0.92-1.1 using the Ramaiah Medical College and Hospitals, Children's Hospital Boston-Massachusetts Institute of Technology, Temple University Hospital, Maastricht University Medical Centre, and University of Bonn databases respectively.

CONCLUSIONS

We observe that the AM-FBC plays a significant role in improving seizure detection results by overcoming inter-database variation of feature distribution.

SIGNIFICANCE

To the best of the author's knowledge, this is the first study reporting on the cross-database evaluation of classification of epileptic seizures and proven to be better generalization capability when evaluated using five databases and can contribute to accurate and robust detection of epileptic seizures in real-time.

Real-Time Localization of Epileptogenic Foci EEG Signals: An FPGA-Based Implementation

The epileptogenic focus is a brain area that may be surgically removed to control of epileptic seizures. Locating it is an essential and crucial step prior to the surgical treatment. However, given the difficulty of determining the localization of this brain region responsible of the initial seizure discharge, many works have proposed machine learning methods for the automatic classification of focal and non-focal electroencephalographic (EEG) signals. These works use automatic classification as an analysis tool for helping neurosurgeons to identify focal areas off-line, out of surgery, during the processing of the huge amount of information collected during several days of patient monitoring. In turn, this paper proposes an automatic classification procedure capable of assisting neurosurgeons online, during the resective epilepsy surgery, to refine the localization of the epileptogenic area to be resected, if they have doubts. This goal requires a real-time implementation with as low a computational cost as possible. For that reason, this work proposes both a feature set and a classifier model that minimizes the computational load while preserving the classification accuracy at 95.5%, a level similar to previous works. In addition, the classification procedure has been implemented on a FPGA device to determine its resource needs and throughput. Thus, it can be concluded that such a device can embed the whole classification process, from accepting raw signals to the delivery of the classification results in a cost-effective Xilinx Spartan-6 FPGA device. This real-time implementation begins providing results after a 5 s latency, and later, can deliver floating-point classification results at 3.5 Hz rate, using overlapped time-windows.

Sleep apnea detection from ECG using variational mode decomposition

Sleep apnea is a pervasive breathing problem during night sleep, and its repetitive occurrence causes various health problems. Polysomnography is commonly used for apnea screening which is an expensive, time-consuming, and complex process. In this paper, a simple but efficient technique based on the variational mode decomposition (VMD) for automated detection of sleep apnea from single-lead ECG is proposed. The heart rate variability and ECG-derived respiration signals obtained from ECG are decomposed into different modes using the VMD, and these modes are used for extracting different features including spectral entropies, interquartile range, and energy. The principal component analysis is employed to reduce the dimension of the feature vector. The experiments are conducted using the Apnea-ECG dataset, and the classification performance of various classifiers is investigated. In per-segment classification, an accuracy of about 87.5% (Sens: 84.9%, Spec: 88.2%) is achieved using the K-nearest neighbor classifier. In per-recording classification, the proposed technique using the linear discriminant analysis model outperformed the existing apnea detection approaches by achieving the accuracy of 100%. The algorithm also provided the best agreement between the estimated and reference apnea-hypopnea index (AHI) values. These results show that the algorithm has the potential to be used for home-based apnea screening systems.

Adaptive-Cognitive Kalman Filter and Neural Network for an Upgraded Nondispersive Thermopile Device to Detect and Analyze Fusarium Spores

Noises such as thermal noise, background noise or burst noise can reduce the reliability and confidence of measurement devices. In this work, a recursive and adaptive Kalman filter is proposed to detect and process burst noise or outliers and thermal noise, which are popular in electrical and electronic devices. The Kalman filter and neural network are used to preprocess data of three detectors of a nondispersive thermopile device, which is used to detect and quantify Fusarium spores. The detectors are broadband (1 µm to 20 µm), λ 1 (6.09 ± 0.06 µm) and λ 2 (9.49 ± 0.44 µm) thermopiles. Additionally, an artificial neural network (NN) is applied to process background noise effects. The adaptive and cognitive Kalman Filter helps to improve the training time of the neural network and the absolute error of the thermopile data. Without applying the Kalman filter for λ 1 thermopile, it took 12 min 09 s to train the NN and reach the absolute error of 2.7453 × 104 (n. u.). With the Kalman filter, it took 46 s to train the NN to reach the absolute error of 1.4374 × 104 (n. u.) for λ 1 thermopile. Similarly, to the λ 2 (9.49 ± 0.44 µm) thermopile, the training improved from 9 min 13 s to 1 min and the absolute error of 2.3999 × 105 (n. u.) to the absolute error of 1.76485 × 105 (n. u.) respectively. The three-thermopile system has proven that it can improve the reliability in detection of Fusarium spores by adding the broadband thermopile. The method developed in this work can be employed for devices that encounter similar noise problems.

EEG-based single-channel authentication systems with optimum electrode placement for different mental activities

BACKGROUND

Electroencephalogram (EEG) signals of a brain contain a unique pattern for each person and the potential for biometric applications. Authentication and security is a very important issue in our life and brainwave-based authentication is an addition to biometric authentication systems, which has many advantages over others. In this paper, we study the performance of a single channel brainwave-based authentication systems and select optimum channels based on mental activities.

METHODS

In this study, we used a dataset with five mental activities with seven subjects (325 samples). The EEG based authentication system includes three pre-processing steps, feature extraction, and classification. Features for Subject Authentication, are obtained from discrete Fourier transform, discrete wavelet transform, autoregressive modeling, and entropy features. Then these features are classified using the Neural Network, Bayesian network and Support Vector Machine.

RESULTS

We achieved accuracy in the range of 97-98% mean accuracy with Neural Network classifier for single-channel authentication system with optimum electrode placement for mental activity. We also analyzed the authentication system independently from the type of mental activity and chose channel O₂ as the optimum channel with an accuracy of 95%.

CONCLUSIONS

Channel optimization can obtain higher performance by reducing the number of EEG channels and defined the optimum electrode placement for different mental activities.

Pozitif ve Negatif Duyguların Ayrımında Etkili EEG Kanallarının Dalgacık Dönüşümü ve Destek Vektör Makineleri ile Belirlenmesi

Duygular kişilerin yaşamlarını ve karar verme mekanizmalarını hayatının tamamında etkilemektedir. İnsanlar duygulara kelimeleri, sesleri, yüz mimiklerini ve vücut dillerini kullanarak istemli ya istemsiz bir şekilde, iş yaparken, gözlemlerken, düşünürken kısacası çevresiyle iletişim kurarken başvururlar. Bundan dolayı, duyguların davranışlarını analiz etmek ve anlamak büyük önem arz etmektedir. Beyin sinyallerine dayalı gerçekleştirilen duygu tahmini günümüzde Beyin-Bilgisayar Arayüzü (BBA) uygulamalarında büyük yarar sağlamaktadır. BBA uygulamaları daha çok sağlık, eğitim, güvenlik, sanal gerçeklik, bilgisayar oyunları olmak üzere birbirinden farklı birçok alanda kullanılmaktadır. Ancak, beyin sinyallerinin elde edilmesi sırasında gürültülerin ortaya çıkması, EEG kanallarının yanlış seçilmesi, verilerin yoğun olması ve uygun olmayan özellik çıkarım yöntemlerinin kullanılması, BBA uygulamalarının yeterli seviyeye gelememelerine neden olmaktadır. Bu çalışmada, hangi EEG kanallarının pozitif-negatif duyguların ayrımında etkili olduğu belirlenmeye çalışılmış ve DEAP veri setindeki 32 kanallı EEG sinyalleri kullanılmıştır. Özellik çıkarım aşamasında, dalgacık dönüşümü, bilgi ölçüm yöntemleri ve istatistiksel yöntemler kullanılarak etkili EEG kanallarının belirlenmesi hedeflenmiştir. Çalışmanın son aşamasında ise, elde edilen özelliklerden yola çıkılarak oluşturulan eğitim kümesi DVM (Destek Vektör Makineleri) kullanılarak sınıflandırılmıştır. Önerilen yöntemin sınıflandırma performansı, sınıflandırma kesinliği, log-kaybı ve on kat çapraz-doğrulama) ile belirlenmiştir. Her bir EEG kanalı için doğruluk oranı hesaplanmış ve ortalama başarım %74 olacak şekilde gözlemlenmiştir. Önerilen yöntem ve tekniklere göre en etkili EEG kanalları Fp1, FC6, C4, CP1, CP5, CP6, T7, P7 ve Pz olarak belirlenmiştir.

Dynamic Hilbert warping, a new measure of RR-interval signals evaluated in the cognitive load estimation

RR interval (RRI) signals represent the time intervals between successive heart R-waves. These signals are influenced by many cognitive and psychological processes. In this study, a new technique based on the combination of empirical mode decomposition and dynamic Hilbert warping (DHW) was proposed to inference cognitive states from measured RRI signals. Moreover, a set of entropic and statistical measures was extracted to characterize the regularity and temporal distribution in the phase spectra and amplitude envelope of the analytic signals. The discriminating capability of the proposed method was studied in 45 healthy subjects. They performed an arithmetic task with five levels of difficulty. The study indicated the importance of phase information in cognitive load estimation (CLE). The new phase characteristics were able to extract hidden information from the RRI signals. The results revealed a striking decrease in DHW value with increasing load level. The entropic measures of analytic signal also showed an increasing trend as the mental load increased. Although, phase information had an ability to discriminate between more distinct levels as well as between more similar ones, amplitude information was effective only in discriminating between more distinct levels.

Classification of epileptic seizures using wavelet packet log energy and norm entropies with recurrent Elman neural network classifier

Electroencephalogram shortly termed as EEG is considered as the fundamental segment for the assessment of the neural activities in the brain. In cognitive neuroscience domain, EEG-based assessment method is found to be superior due to its non-invasive ability to detect deep brain structure while exhibiting superior spatial resolutions. Especially for studying the neurodynamic behavior of epileptic seizures, EEG recordings reflect the neuronal activity of the brain and thus provide required clinical diagnostic information for the neurologist. This specific proposed study makes use of wavelet packet based log and norm entropies with a recurrent Elman neural network (REN) for the automated detection of epileptic seizures. Three conditions, normal, pre-ictal and epileptic EEG recordings were considered for the proposed study. An adaptive Weiner filter was initially applied to remove the power line noise of 50 Hz from raw EEG recordings. Raw EEGs were segmented into 1 s patterns to ensure stationarity of the signal. Then wavelet packet using Haar wavelet with a five level decomposition was introduced and two entropies, log and norm were estimated and were applied to REN classifier to perform binary classification. The non-linear Wilcoxon statistical test was applied to observe the variation in the features under these conditions. The effect of log energy entropy (without wavelets) was also studied. It was found from the simulation results that the wavelet packet log entropy with REN classifier yielded a classification accuracy of 99.70 % for normal-pre-ictal, 99.70 % for normal-epileptic and 99.85 % for pre-ictal-epileptic.

Classification of Obsessive Compulsive Disorder by EEG Complexity and Hemispheric Dependency Measurements

In the present study, both single channel electroencephalography (EEG) complexity and two channel interhemispheric dependency measurements have newly been examined for classification of patients with obsessive–compulsive disorder (OCD) and controls by using support vector machine classifiers. Three embedding entropy measurements (approximate entropy, sample entropy, permutation entropy (PermEn)) are used to estimate single channel EEG complexity for 19-channel eyes closed cortical measurements. Mean coherence and mutual information are examined to measure the level of interhemispheric dependency in frequency and statistical domain, respectively for eight distinct electrode pairs placed on the scalp with respect to the international 10–20 electrode placement system. All methods are applied to short EEG segments of 2 s. The classification performance is measured 20 times with different 2-fold cross-validation data for both single channel complexity features (19 features) and interhemispheric dependency features (eight features). The highest classification accuracy of 85 ±5.2% is provided by PermEn at prefrontal regions of the brain. Even if the classification success do not provided by other methods as high as PermEn, the clear differences between patients and controls at prefrontal regions can also be obtained by using other methods except coherence. In conclusion, OCD, defined as illness of orbitofronto-striatal structures [Beucke et al., JAMA Psychiatry 70 (2013) 619–629; Cavedini et al., Psychiatry Res. 78 (1998) 21–28; Menzies et al., Neurosci. Biobehav. Rev. 32(3) (2008) 525–549], is caused by functional abnormalities in the pre-frontal regions. Particularly, patients are characterized by lower EEG complexity at both pre-frontal regions and right fronto-temporal locations. Our results are compatible with imaging studies that define OCD as a sub group of anxiety disorders exhibited a decreased complexity (such as anorexia nervosa [Toth et al., Int. J. Psychophysiol. 51(3) (2004) 253–260] and panic disorder [Bob et al., Physiol. Res. 55 (2006) S113–S119]).

Controlling a virtual forehand prosthesis using an adaptive and affective Human-Machine Interface

This paper presents the design of an adaptable Human-Machine Interface (HMI) for controlling virtual forearm prosthesis. Direct physical performance measures (obtained score and completion time) for the requested tasks were calculated. Furthermore, bioelectric signals from the forehead were recorded using one pair of electrodes placed on the frontal region of the subject head to extract the mental (affective) measures while performing the tasks. By employing the proposed algorithm and above measures, the proposed HMI can adapt itself to the subject's mental states, thus improving the usability of the interface. The quantitative results from 15 subjects show that the proposed HMI achieved better physical performance measures in comparison to a conventional non-adaptive myoelectric controller (p < 0.001).