Please use this identifier to cite or link to this item:
Title: Improved EEG Signal Analysis Techniques for Epileptic Spike Detection and Artifact Excision
Authors: Garg, Harish Kumar
Kohli, Amit Kumar (Guide)
Keywords: EEG
Epileptic Spikes
Issue Date: 2-Sep-2017
Abstract: Electroencephalography is the most common noninvasive technique utilized for monitoring the electrical brain activity and inferring brain function, in which the recorded signal traces represent an electric signal from a large number of neurons. The main goal of electroencephalogram (EEG) signal analysis is to infer functional connectivity between different brain areas, which is directly useful for neuroscience and clinical investigations. Due to its potentially complex nature and due to the presence of epileptic-spikes (ESs), ocular-artifacts (OAs) and inevitable additive-white-Gaussian-noise (AWGN), the electroencephalogram signal processing poses some great challenges for researchers. These challenges can be tackled, by using the epileptic spike detection, artifact suppression and noise removal techniques, in a principled manner via adaptive signal processing approach. We first present the evaluation of nonstationary epileptic spike detection algorithm for the electroencephalogram signal using the smoothed-nonlinear-energy-operator (SNEO) based on the different time-domain window functions. However, the incorporation of adaptive threshold determination procedure enhances the performance of proposed ES detector. The detection procedure exploits the fact that the presence of instantaneous ES corresponds to the high instantaneous energy content at the high frequencies. In addition to the stochastic amplitude, sign and the location of appearance of triangular spikes in the synthetic EEG signal, its base-width is also considered to be variable for the nonstationary signal analysis. The five pairs of EEG signals, obtained from electrodes placed on the left and right frontal cortex of male adult WAG/Rij rats, are used for the testing of proposed adaptive scheme in the real-time environment, which is a genetic animal model of human epilepsy. The simulation results are presented to demonstrate that the choice of window function plays a pivotal role in the efficient detection of ESs. Its computational complexity is found to be in trade-off relationship with the detection accuracy of algorithm. However, it may be inferred that the real-time EEG signals (rat-data) can be processed and analyzed using the proposed adaptive scheme for the ES detection, which supersedes the conventional techniques. We next propose a technique for the electroencephalogram spike enhancement and detection, which uses the Kalman-filtering (KF) approach based on the output correlation method for the nonstationary signal enhancement. We describe the nonstationary EEG signal in terms of the general Markov-model, in which the parameters are considered to be time-varying. In the proposed methodology, neither the process- and measurement-noise statistics nor the initial Kalman blending factor are stringently required. The EEG epileptic spikes are pre-emphasized using the output correlation method, and subsequently the detection is performed using the decision threshold based on the output of same adaptive filter. We have tested the proposed scheme on the synthetic EEG signal corrupted with randomly occurring triangular spikes. The presented simulation results manifest significant improvement in the signal-to-noise-ratio (SNR) due to the modified estimation of time-varying parameters of the general Markov-model, which in turn leads to the alleviated number of false-positives (FPs). It is apparent that the real-time EEG signal (rat-data) can be analyzed using the proposed EEG epileptic spike enhancement and detection adaptive scheme, which outperforms the conventional KF technique under the different SNR conditions. At 10dB SNR, the output correlation method provides approximately 40 % reduction in FPs for the triangular spikes in synthetic EEG signal and approximately 27.5 % reduction in FPs for ESs in the rat-data as compared to the conventional KF scheme. Further, we present a technique for the removal of ocular artifacts from the electroencephalogram by using the adaptive filtering. The major concern is electrooculogram (EOG) signal present in the recorded EEG signal, which appears due to the abrupt eye movements. The conventional regression based methods for removing the EOG artifacts require various procedures for the pre-processing and calibration, which are inconvenient and time-consuming. However in the presented method, we use separately recorded horizontal-EOG (HEOG) and vertical-EOG (VEOG) signals as two reference inputs, which are processed using the finite-impulse-response (FIR) filters. The linear filter coefficients are adaptively updated using the numeric-variable-forgetting-factor (NVFF) recursive-least-squares (RLS) algorithm, which track the nonstationary EOG signals. Subsequently, the processed HEOG and VEOG signals are subtracted from the recorded EEG signal to obtain the artifact-free EEG signal. The most appealing feature of this artifact and noise excision technique is its ability to remove the EOG artifacts without any pre-processing and calibration. The weight-vector of FIR filters can be automatically adapted to a new state without losing its effectiveness. Simulation is conducted using the synthetic EEG signal corrupted by the noise, synthetic HEOG and VEOG signals. The real-time recorded EEG signal (corrupted by EOG and noise) is also refined using the separately recorded reference EOG signals and an FIR filter. The linear generalized-variable-step-size normalized-least-mean-square (GVSS-NLMS) algorithm based artifact and noise suppression scheme is found to be suitable only under the high signal-to-noise-ratio conditions, due to its low computational complexity. However for the synthetic as well as real-time signals, the simulation results are presented to demonstrate that the linear NVFF-RLS algorithm based artifact and noise excision technique outperforms the conventional fixed-forgetting-factor (FFF) RLS, fixed-step-size (FSS) NLMS and GVSS-NLMS algorithms, in terms of the reduction in mean squared error, under the low as well as high SNR conditions. In the presented research work, we are incorporating different adaptive algorithms to formulate improved EEG signal analysis techniques for the epileptic spike detection and artifact/noise excision, which only represents the tip of the iceberg in the domain of biosignal processing and biomedical engineering.
Description: PhD Dissertation
Appears in Collections:Doctoral Theses@ECED

Files in This Item:
File Description SizeFormat 
PhD_Thesis_HarishKrGarg_Sept2017.pdfPhD Thesis 2017982.9 kBAdobe PDFView/Open    Request a copy

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.