Research

Article: Capsule neural network based approach for subject specific and cross-subjects seizure detection from EEG signals

Abstract: The objective of this study is to propose an approach to detect Seizure and Non-Seizure phenomenon from the highly inconsistent and non-linear EEG signals. In the view of performing cross-subject classification over the inconsistency and non-linear characteristics of EEG signals, we have proposed a fine-tuned Capsule Neural Network (CapsNet) based approach to classify the seizure and non-seizure EEG signals through subject specific and cross-subject training and testing. In this experiment, first we have normalized the input data using L2 normalization technique. In the second step, the normalized data have been given to the CapsNet and model level fine-tuning has been carried out. In addition to this, we have performed seizure and non-seizure classification performance evaluation using three more classifiers such as Decision Tree, Logistic Regression, Convolutional Neural Network to compare with the performance of the proposed approach. To estimate the effectiveness of the proposed approach, subject specific and cross-subject training and testing have been performed. In both experiments, we have used multi-channel and single channel EEG datasets. For subject specific experiment, the proposed approach achieved a mean accuracy of 93.50% over the dataset-1 (multichannel) and an accuracy of 82.61% for dataset-2 (single channel). For cross-subject experiment, the proposed approach achieved a highest mean accuracy of 86.41% over the dataset-1(multi-channel) and a mean accuracy of 48.45% over the dataset-2 (single channel) which shows an advantage of CapsNet in a certain data scenario as described in result section. Overall performance of the proposed approach shown a comparable improvement over the existing approaches.

Keywords: Electroencephalogram (EEG); Cross-subject seizure detection; Capsule neural network; Decision tree; Logistic regression; Convolutional neural network.

Article Link: DOI

Article: DWT-EMD Feature Level Fusion Based Approach over Multi and Single Channel EEG Signals for Seizure Detection

Abstract: Brain Computer Interface technology enables a pathway for analyzing EEG signals for seizure detection. EEG signal decomposition, features extraction and machine learning techniques are more familiar in seizure detection. However, selecting decomposition technique and concatenation of their features for seizure detection is still in the state-of-the-art phase. This work proposes DWT-EMD Feature level Fusion-based seizure detection approach over multi and single channel EEG signals and studied the usability of discrete wavelet transform (DWT) and empirical mode decomposition (EMD) feature fusion with respect to individual DWT and EMD features over classifiers SVM, SVM with RBF kernel, decision tree and bagging classifier for seizure detection. All classifiers achieved an improved performance over DWT-EMD feature level fusion for two benchmark seizure detection EEG datasets. Detailed quantification results have been mentioned in the Results section.

Keywords: Discrete wavelet transform; empirical mode decomposition; electroencephalogram; EEG classification; seizure detection

Article Link: DOI

Article: Capsule neural networks on spatio-temporal EEG frames for cross-subject emotion recognition

Abstract: Scalp EEG plots are plots of scalp potentials against time, and hence, capture spatial information, owing to the placement of electrodes on the scalp, as well as, temporal information from variations in brain waves. In this paper we propose a novel method to make a combined representation of spatial and temporal information, by incorporating the signals into a sparse spatio-temporal frame, such that it can be easily processed by deep learning algorithms in the computer vision domain. Familiarities of a model to the test data in the setting of emotion recognition from EEG, is also defined, and a form of data splitting such that the model has to perform on a set with which it has the minimum degree of familiarity is introduced. A CapsNet architecture is trained on DEAP dataset to perform on a cross-subject binary classification task, and tuning of the hyperparameters using Bayesian Optimization is analyzed. The proposed model reports a best-case accuracy of 0.85396 and average case accuracy of 0.57165 for LOO subject, and a best case of 1.0 and average case of 0.51071 for unseen-subjectunseen-record classification, when averaged across all the classes (i.e., valence, dominance, arousal, and liking), which is comparable to that reported by other works.

Keywords: Spatio-temporal representation; Spatio-temporal EEG; CapsNet; EEG Emotion Recognition.

Article Link: DOI

Article: A Multi-View SVM Approach for Seizure Detection from Single Channel EEG Signals

Abstract: Seizures are the part of the epilepsy that occurs in central nervous system which leads abnormal brain activity. Electroencephalogram (EEG) signal recordings are mostly used in epileptic seizure detection process. Detection of seizures is a crucial part for further treatment of patients. This paper proposes a multi-view SVM model for seizure detection using the single channel EEG signals. In this experiment, two views of the EEG data have been extracted, (1) the time domain features using Independent Component Analysis (ICA) and (2) power spectral densities are obtained in the frequency domain. Extracted features have been fed to multi-view SVM classification model. In this study, a single channel EEG dataset is used for seizure detection. Performance estimation parameters namely Accuracy, Sensitivity, Specificity, F1-score, and AUC value have been estimated for evaluating the proposed model. The model classified seizure and non-seizure over the sets A vs E and B vs E with an accuracy greater than 99% using k-fold cross validation. The classification accuracy obtained by multi-view SVM is better by 1–4% than single view SVM using the same features. Furthermore, the proposed model is also compared with existing single view SVM models. It is observed that the multi view SVM model performed significantly better, compare to a single view SVM model over the same features.

Keywords: Electroencephalogram; EEG signals; Epilepsy; Independent Component Analysis (ICA); Power Spectral Density (PSD); Seizure; Support vector machine (SVM).

Article Link: DOI

Article: A 1D-CNN-Spectrogram Based Approach for Seizure Detection from EEG Signal

Abstract: This work proposes a scheme for Seizure Detection from the Electroencephalogram (EEG) signal data generated due to the electrochemical process in the human nervous system and acquired from the human brain scalp using electrode placement. In this experiment, the complete workflow divided into three parts to get a better performance of our proposed methodology. These Parts are raw EEG Signal data filtering, Spectrogram feature matrix generation and finally, One-Dimensional Convolution Network has been used for Seizure Detection. Thus, our main objective in this work is to represent a methodology with the combination of two methods Spectrogram and 1D CNN which can be one possible approach for seizure detection. Our proposed methodology achieved comparable performance in terms of Sensitivity, Specificity, and Accuracy.

Keywords: Convolution Neural Network (CNN); Seizure; Electroencephalogram; Spectrogram; Sensitivity; Specificity; Accuracy.

Article Link: DOI

Article: EEG Channel Selection Approach for Seizure Detection Based on Integrated BPSO and ELM

Abstract: Seizure is an unexpected, uncontrollable disturbance in the neuronal activity of the brain in which human being encounters changes in functioning, motion, emotions, and in levels of consciousness. The paper proposes a methodology for the detection of seizures using Electroencephalogram (EEG) signals through Channel Selection. Seizure detection using all the EEG channels leads less detection accuracy as well as high computation. The trade-off between accuracy and computation is done by using channel optimization techniques. The proposed methodology consists of a channel optimization technique which integrates Binary Particle Swarm Optimization (BPSO) with Extreme Learning Machine (ELM) to give the optimized set of channels. The data of the optimized set of channels is then classified using Extreme Learning Machine for gauging its performance. The result obtained from the proposed methodology is compared with the results without channel optimization technique. Also, the result is compared with the combination of channels obtained through maximized and minimized Fitness function. The proposed methodology achieved a maximum detection accuracy of 93.21% with only 6/23 channels, whereas ELM achieved a maximum detection accuracy of 90.27% without channel Selection. Thus, it effectively solves the problem associated with using all the channels and achieves better detection accuracy.

Keywords: Binary Particle Swarm Optimization (BPSO); Extreme Learning Machine (ELM); Electroencephalogram(EEG), Seizure. .

Article Link: DOI

Article: Performance Analysis of Supervised Machine Learning Algorithms for Epileptic Seizure Detection with high variability EEG datasets: A Comparative Study

Abstract: A wide variety of machine learning techniques have been developed which could aid in the analysis of continuous EEG time-series data and derive useful insights from it. The techniques which have been most popularly used for such analysis include Neural Networks, Support Vector Machines, and Linear Discriminant Analysis. Extreme Learning Machines are one of the newest additions to these classifiers. The present work compares the performance of these classifiers on both single channel and multi-channel EEG recordings. Two different datasets used for the experiments are (a) EEG database by University of Bonn, Germany for single channel recordings, (b) CHB-MIT dataset for multi-channel recordings. In all these experiments, we assume that the recordings, after preprocessing, are all free from such defects which could affect the performance of different classifiers differently. For preprocessing, filtered recordings are segmented and DWT is applied on them to use the transform coefficients for feature extraction. These extracted features were then fed into various classifiers. It was observed that ELM classifiers could perform at par or better than the conventional classification methods.

Keywords: Artificial Neural Networks (ANN), Support Vector Machines (SVM), Linear Discriminant Analysis (LDA), Extreme Learning Machines (ELM), Electroencephalogram (EEG), Discrete Wavelet Transform (DWT).

Article Link: DOI

Article: Performance Estimation and Analysis Over the Supervised Learning Approaches for Motor Imagery EEG Signals Classification

Abstract: In this paper, a comparative analysis has been done to estimate a robust classifier to classify motor imagery EEG data. First, segment detection and feature extraction have been done over the rawEEGdata. Then the frequency domain features have been extracted using FFT. Six classifiers DNN, SVM, KNN, Naïve Bayes,’ Random Forest, and Decision Tree have been considered for this study. The DNN model configured with four layers and used the binary cross-entropy loss function and sigmoid activation function for all layers. The optimizer used is “Adam” having the default-learning rate of 0.001. In this experiment, for the purpose of the estimation of the performance of various classifiers, the experiment used dataset IVa from BCI Competition III, which consisted of EEG signal data for five subjects, namely ‘aa,’ ‘al,’ ‘av,’ ‘aw,’ and ‘ay.’ The highest average accuracy of 70.32% achieved by the DNN model, whereas the model achieved an accuracy of 80.39% over the subject ‘aw.’ The objective of this experiment encompasses the different models for the classification of various motor tasks from EEG signals.

Keywords: Electroencephalogram (EEG); Brain–computer interface (BCI); Motor imagery; Deep neural network (DNN); SVM; KNN; Naive Bayes; Random forest; Decision tree. .

Article Link: DOI

Article: A Deep Transfer Learning Approach for Seizure detection using RGB features of Epileptic Electroencephalogram Signals

Abstract: This paper demonstrates an approach based on Deep Transfer Learning for the classification for Seizure and Nonseizure Electroencephalogram (EEG) signals. Recognizing seizure signals in intelligent way is quite important in clinical diagnosis of Epileptic seizure. Various traditional and deep machine learning techniques are employed for this purpose. However, the Epileptic seizure prediction and classification performance is not satisfactory over small EEG dataset using traditional approaches. The Transfer learning approach overcomes this by reusing the pre-trained networks such as googlenet, resnet101 and vgg19 trained on large Image database. This experiment has been done in two phases: (1) RGB image dataset generated for the seizure and non-seizure EEG signals data of University of Bonn using a novel preprocessing technique, (2) we configured googlenet, resnet101 and vgg19 trained networks to learn a new pattern or features from the RGB image Dataset and finally, above mentioned networks have been used for the classification. The use of Vgg19 network shows greater accuracy among the three but takes comparatively more prediction time. We will mainly emphasize on the results obtained from the googlenet, since it provides effective accuracy taking less time for prediction. The proposed method achieved an accuracy of above 99% for a smaller number of epochs and maximum accuracy of 100% when we increase number of epochs. Experimental outcomes show the proposed approach using googlent achieved better performance w.r.t to many state-of-the-art classification algorithms even on the small EEG dataset. In addition, classification performance of our proposed approach has compared with different traditional machine learning techniques over the same input data.

Keywords: Transfer Learning; Electroencephalogram Signal; Epileptic Seizure; Deep Learning; googlenet; resnet101; vgg19.

Article Link: DOI