Access the full text.
Sign up today, get DeepDyve free for 14 days.
Loading next page...
/lp/hindawi-publishing-corporation/source-localization-of-eeg-brainwaves-activities-via-mother-wavelets-pT0vfgiDqm
References
References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.
- Publisher
- Hindawi Publishing Corporation
- Copyright
- Copyright © 2021 Tarek Frikha et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
- ISSN
- 2040-2295
- eISSN
- 2040-2309
- DOI
- 10.1155/2021/9938646
- Publisher site
- See Article on Publisher Site
Abstract
Hindawi Journal of Healthcare Engineering Volume 2021, Article ID 9938646, 11 pages https://doi.org/10.1155/2021/9938646 Research Article Source Localization of EEG Brainwaves Activities via Mother Wavelets Families for SWT Decomposition 1 1 2 3 Tarek Frikha , Najmeddine Abdennour, Faten Chaabane , Oussama Ghorbel, 3 3 4 Rami Ayedi, Osama R. Shahin, and Omar Cheikhrouhou CES Lab, Universit´e de Sfax, Sfax, Tunisia Regim-Lab, Universit´e de Sfax, Sfax 3038, Tunisia Jouf University, Sakakah, Saudi Arabia College of CIT, Taif University, P.O. Box 11099, Taif 21944, Saudi Arabia Correspondence should be addressed to Tarek Frikha; tarek.frikha@enis.tn Received 15 March 2021; Revised 5 April 2021; Accepted 17 April 2021; Published 28 April 2021 Academic Editor: Dr. Dilbag Singh Copyright © 2021 Tarek Frikha et al. )is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. A Brain-Computer Interface (BCI) is a system used to communicate with an external world through the brain activity. )e brain activity is measured by electroencephalography (EEG) signal and then processed by a BCI system. EEG source reconstruction could be a way to improve the accuracy of EEG classification in EEG based brain-computer interface (BCI). )e source localization of the human brain activities can be an important resource for the recognition of the cognitive state, medical disorders, and a better understanding of the brain in general. In this study, we have compared 51 mother wavelets taken from 7 different wavelet families, which are applied to a Stationary Wavelet Transform (SWT) decomposition of an EEG signal. )is process includes Haar, Symlets, Daubechies, Coiflets, Discrete Meyer, Biorthogonal, and reverse Biorthogonal wavelet families in extracting five different brainwave subbands for source localization. For this process, we used the Independent Component Analysis (ICA) for feature extraction followed by the Boundary Element Model (BEM) and the Equivalent Current Dipole (ECD) for the forward and inverse problem solutions. )e evaluation results in investigating the optimal mother wavelet for source localization eventually identified the sym20 mother wavelet as the best choice followed by bior6.8 and coif5. regions of the cerebral cortex are involved in the planning, 1. Introduction control, and execution of voluntary movements. Electro- Brain-Computer Interface (BCI) not only external permits encephalography (EEG) signals are electrical potentials controlling devices but also interacts using the environment generated by the nerve cells in the cerebral cortex. In order to by brain signals. EEG signals measurements over the motor execute motoric tasks, the EEG signals have appeared over cortex exhibit changes in power related to the movements or the motor cortex [1]. To accurately study and analyze the human brain, imaginations which are executed in motor tasks [1]. Changes declare decrease or increase of power in alpha (8 Hz–13 Hz) electroencephalography (EEG) [1] is thought to be the and beta (13 Hz–28 Hz) frequency bands from resting state optimal method that helps us advance in our quest due to the to motor imagery task known as event related synchro- noninvasiveness and the low-cost factors. )e electroen- nization and desynchronization [2]. )e necessity to com- cephalogram (EEG) is a recording of the electrical activity of municate with the external world for locked-in state (LIS) the brain from the scalp. )e recorded waveforms reflect the patients made doctors and engineers motivated to develop a cortical electrical activity. In fact, the EEG provides access to BCI technology for typing letters through brain commands. human brain activities in 5 main frequency band packages Research has been done around this area to ascertain the presented in Table 1 [2]. )e scientific trend shifted towards dream of typing for the handicapped. In the brain, some exploiting these frequency subbands and seeking the 2 Journal of Healthcare Engineering Table 1: )e EEG signal main human brain frequencies. EEG bands Frequency (Hz) Main description Delta 0–4 Deep state of sleep )eta 4–8 Deep meditation and lucid dreaming Alpha 8–12 Relaxation/creativity Beta 12–32 Analytical thinking or stress/anxiety Lower 32–64 Gamma Wide brain activities or higher 64–128 brain disorder Higher 64–128 extraction of pure and noncontaminated signals instead of 2. Methodology developing recording methods and other ways to express the 2.1. Dataset signal. )e research community took an omnidirectional ap- 2.1.1. Simulated Signal. Influenced by the morphology and proach throughout the recent years to try to extract the the structure of actual EEG signals, we created a sinusoidal human brain activities and access these five different fre- signal with oscillations of 400 ms on 800 ms time windows quency subbands. In this context, Murali et al. [3] used the used in the evaluation process. recurrence quantification analysis (RQA) algorithm and an A sampling rate of 1000 Hz and an oscillation frequency adaptive FIR filter for the EEG signal extraction. As for of 3, 6, 10, 20, and 45 Hz were recorded for the extraction of Singh, Vivek et al. [4], they compared using Finite Impulse Delta, )eta, Alpha, Beta, and Gamma waves, respectively. Response (FIR) and Infinite Impulse Response (IIR) filters )e signal to the noise ratio (SNR) was also altered from −5 and confirmed the FIR success over RII regarding EEG to 15 dB with −5 dB for noisy signal simulation, 10 dB for signals. On the other hand, Nallamothu et al. [5] used a balanced signals, and 15 dB for acceptable quality signals. On Nonlinear Least Mean square (LMS) adaptive filtering to the other hand, the amplitude of the signal depends on both remove artifacts from the EEG signal. the SNR value and the noise contaminating the signal, which For their part, Tzimourta et al. [6, 7] used a Discrete is what is known as a pink noise; besides, it is a very common Wavelet Transform (DWT) for feature extraction and noise for biological systems. Support Vector Machine (SVM) classification of Epileptic Seizures. Actually, our previous work [8] has proved the effectiveness of the Stationary Wavelet Transform (SWT) 2.1.2. /e EEG Dataset. )e EEG signal dataset used in this using Symlet 4 mother wavelet compared to FIR filters in study is a one-subject recording of a presurgical EEG signal feature extraction of the alpha and gamma band waves. from a pharmacoresistant subject with asymptomatic focal Furthermore, Akkar and Jasim [9] proved that the Symlet 9 cortical dysplasia in the right occipital-temporal junction. mother wavelet is the best wavelet from a set of 25 mother )e acquisition and preprocessing phases were applied as in wavelet functions using the Packet Wavelet Transform our previous work [7, 8] and validated by an expert neu- (PWT). Condo and Efren ´ [10] compared 18 different mother rologist. )is particular EEG recording was chosen because wavelets for EEG signal analysis and affirmed Symlet 6 and it presented clear alpha and gamma patterns with regular Daubechie 5 are the most adequate for EEG signals. Noor spiking and visible epileptic oscillations as validated by the et al. [11] compared 45 mother wavelets to conclude that expert. )e EEG data was recorded on a Deltamed System, Symlet 9 followed by Coiflet 3 and Daubechie 7 exhibits the with a 2500 Hz sampling rate and antialiasing low-pass highest similarities and compatibilities with the EEG signal analog filter set to 100 Hz. )e dataset contained 74 epochs after applying an FIR notch filter. with a 6-second duration each, 62 channels, and 148 events. )e EEG signal is a nonstationary signal; the advantage of using the wavelet transform over the usual Fourier 2.2. /e Wavelet Transform. Similar to the Fourier transform transform in EEG signals is their capability to analyze (FT), the wavelet transform (WT) is a function that grants nonstationary signals [12, 13] due to their improved pre- the passage from the time to the frequency domain. How- sentation in both the time and frequency domain as shown ever, the FT decomposes the signal into a series of sinus and by Figure 1. cosines components as in the following equation: In this context, this study aims at comparing 51 different +∞ mother wavelets using SWT to extract human brainwaves jωt s(t) � (1) S(ω)e dω, and localize their sources. In section 2, we will address the ω�−∞ methodology first, by describing the manipulated dataset with S(ω) the short-time Fourier coefficient controlled using and then proceed by presenting the SWT and the processing the frequency parameter ω. steps of our study and finally by introducing the evaluation )e wavelet transform also decomposes the signal into a methods. Section 3 will feature the conceived results and series of wavelet component as in the following equation: section 4 will highlight the discussion. Journal of Healthcare Engineering 3 Time Time Fourier transform Wavelet transform (a) (b) Figure 1: Comparison between the partition of Fourier transform and wavelet transform in the time-frequency domain. (a) Fourier transform. (b) Wavelet transform. +∞ +∞ As the wavelet decomposition phase is completed, we s(t) � c(a, b)φ (t)da.db, (2) a,b evaluate the mother wavelets used in this process and move a�0 b�0 on to the source localization. Figure 4 shows the processing where C(a, b) is the wavelet coefficient and φ (t) the steps of this study. a,b mother wavelet with “a” the scaling parameter and “b” the wavelet shifting parameter that determines the shape of the wavelet. In fact, Figure 2 highlights the difference between 2.4. /e Evaluation Methods FT and WT decomposition components. Moreover, the 2.4.1. /e Goodness of Fit (GOF). )e goodness of fit (GOF) wavelets are characterized by a limited duration, irregularity, is an evaluation method commonly used for physiological and asymmetricity compared to the predictable, fluid, and signals that adopt Pearson’s chi-squared statistical test [17], infinitely propagated sinus waveform. which is the normalized sum of squared deviations that On the other hand, the wavelet transform used in this investigate the likelihood of an observed difference in the study is the stationary one (SWT) instead of the Continuous frequency distribution compared to the theoretical distri- Wavelet Transform (CWT) or the Discrete Wavelet bution as in the following equation: Transform (DWT). In fact, the SWT is more suitable for our case by avoiding the frequency band overlapping of CWT r r 2 s(t) − s (t) t�1 f [14] and preserving the properties of the signal by averting GOF � 1 – (3) , r 2 s(t) the binary decimation process (downsampling) of DWT t�1 [15, 16]. where s(t) is the theoretical power and s (t) the power of the extracted signal that depends on the adopted mother wavelet. 2.3. Levels of Decomposition and Processing Steps. In order to decompose the EEG signal of our dataset that has 2500 Hz sampling rate to extract the five EEG frequency subbands, 2.4.2. /e Power Spectral Density (PSD) and Scalp Topographies. we had to reduce the signal to exactly 2048 Hz sampling )e Power Spectral Density is a display of the data energy rate; otherwise, these subbands would be extremely distribution throughout the frequency spectrum. It is overlapping. In Figure 3, we display the decomposition of used as a visual evaluation process for its efficiency in the resampled EEG signal. We notice here that, in our presenting the data in the frequency domain rather than previous study [8], we have not resampled the signal as we the time domain, which allows the identification of the extracted only the alpha and gamma waves that were far extracted EEG frequency bands [18]. )e energy fre- separated and did not cause band overlapping issues. Our quency distribution of the EEG signal channels compares decomposition level was 9 to acquire access to the delta the mother wavelets effectiveness in isolating the extracted wave frequencies while our previous work needed only 7 frequency band from the other subbands or artifacts and levels of decomposition to reach the alpha wave. We can differentiates its capabilities to amplify the extracted signal also notice that, in our previous work, the approximated power. coefficients cAi included upper and lower levels (for alpha On the other hand, the scalp topographies are another wave extraction, the cAi were 6, 7, 8 and cDi was 7), while visual evaluation process since it represents a mapping of the for this study, we have included only the above upper levels brain activities distributed on the surface of the scalp. An for the cAi (for alpha wave extraction, the cAi were 6, 7 and increasingly dipolar topography suggests that a cerebral cDi was 7). )e most studied characteristic of EEG signals measurement is an observation of a discharge operation in accordance with alertness level is Power Spectral involving a big number of neurons. Even in nonepileptic Density (PSD) of different brain waves: delta, theta, alpha, observations of brain activities, the dipolar scalp topogra- and beta. phies are a great indicator of a valuable recording session Frequency Frequency 4 Journal of Healthcare Engineering 51 mother wavelet for SWT decomposition EEG signal acquisition (EEG dataset) (a) (b) Source localization of the EEG human brainwaves activities via all the different mother wavelets families for stationary wavelet transform Source localization decomposition with BEM and ICA for ECD (c) (d) Figure 2: Comparison between (a) FT decomposition component and different mother wavelets families decomposition components. Scalp topographies PSD GOF (b) Symlets 4 (c) Coiflets 5. (d) Daubechies 11. Evaluation of the mother wavelets EEG Figure 4: )e cycle of processing steps during this study. signal [0 Hz, 2048 Hz] cA1 cD1 conductivity [21]. For the forward problem, we used the Boundary Element Model (BEM), which is a surface mesh [0 Hz, 1024 Hz] [512 Hz, 1024 Hz] calculation of interfaces between the tissues using the MRI of cA2 cD2 the patient (which makes it a realistic model) [22]. For the [0 Hz, 512 Hz] [256 Hz, 512 Hz] inverse problem, which is an estimation of the current cA3 cD3 generator distribution responsible for the electric EEG [0 Hz, 256 Hz] [128 Hz, 256 Hz] signal, we used the Equivalent Current Dipole (ECD), which cA4 cD4 is the most used method to simplify the brain activities in a [0 Hz, 128 Hz] [64 Hz, 128 Hz] few sources [23]. )e signals are assumed to be generated by cA5 cD5 a small number of focal sources modeled by current dipoles [0 Hz, 64 Hz] [32 Hz, 64 Hz] (an unknown position, amplitude, and orientation). cA6 cD6 Moreover, the extracted signal has to undergo an inde- [0 Hz, 32 Hz] [16 Hz, 32 Hz] pendent component analysis (ICA) dipole fitting operation cA7 cD7 as a preprocessing phase before the ECD inverse problem [0 Hz, 16 Hz] [8 Hz, 16 Hz] solution, in order to separate different components and cA8 cD8 make the components in a dipolar state useful in the lo- [0 Hz, 8 Hz] [4 Hz, 8 Hz] calization of the source generators. )e ICA is the feature cA9 cD9 extraction phase compatible with the statistically indepen- [0 Hz, 4 Hz] [2 Hz, 4 Hz] dent and non-Gaussian signals, which are the traits of the EEG signal [24] while the ECD and the BEM are our Figure 3: EEG signal SWT decomposition levels with cAi as the classification algorithm [25]. approximated coefficients and cDi as the detailed coefficients. In fact, the source localization process is sensitive to the quality of the extracted EEG frequency band and can also since they reflect the domination of certain areas over others serve as an evaluation process that depends on the number of in the energetic exertion, which is the typical and more the located sources and the accuracy of their localization. natural habit of cerebral behavior [19]. 3. Results 2.4.3. /e Source Localization. )e source localization is an 3.1. /e GOF Evaluation Results. )e goodness of fit (GOF) estimation of the brain activity generator locations [20]. To is the evaluation process that enabled us to minimize both reach this estimation, first, we solve the forward problem, our wavelet selection and processing criteria. Considering which is a calculation of the field generated by a given source that the other evaluation methods and the source localiza- for an estimated brain shape and conductivity, with a tion are a computationally heavy and costly process, the consideration of numerous properties, such as the shape of GOF is an excellent fast evaluation that relieved us from the brain that changes from a subject to another or the repeating the hull processing steps and source localization anisotropy conductivity of the skull and the brain for the vast number of 51 mother wavelets. Figure 5 presents Journal of Healthcare Engineering 5 GOF extraction of alpha and gamma waves with 51 mother wavelets evaluation method that grants us a visual representation of the EEG signal extraction. )e choice of frequency subband extraction visualization for this evaluation was limited to the alpha and gamma waves for the confirmed potential of SWT in their extraction. Moreover, due to the weak energy of the gamma wave and its proximity to the 50 Hz noise artifact of the original EEG signal dataset, we relied only on the alpha wave in the PSD visualization as it provides a clear display of the extraction effectiveness difference between the selected mother wavelets. In Figure 8, we compare the EEG signal extraction of the alpha frequency subband using the different noteworthy wavelets chosen by the GOF evaluation ordered from the worst to the best. As we can deduce, the haar and sym4 wavelets, respectively, had the worst results with a signal spectrum contaminated by different artifacts and other frequency subbands while sym20 and dmey had the best Alpha-5 SNR Gamma-5 SNR Alpha-10 SNR Gamma-10 SNR results in isolating the extracted signals from other infil- Alpha-15 SNR Gamma-15 SNR trating ones. We can also recognize the abilities of the new Figure 5: )e GOF results in alpha and gamma waves with 51 SWT decomposition in eliminating high frequency, while mother wavelets using SNR values of −5, 10, and 15 dB. witnessing some difficulties in low-frequency elimination, such as delta and theta, as demonstrated in the PSD visualization. the GOF results for the 51 mother wavelets with different For the scalp topography visualization, almost all the SNR values of −5, 10, and 15 dB as we have mentioned in the noteworthy mother wavelets selected by the GOF had dataset descriptions in Section 2, A, 1). similar good results by producing depolarized scalp to- On the other hand, the use of alpha and gamma wave pographies isolated from the other frequencies, except for extraction in GOF evaluation is justified by our earlier the Haar and sym4 wavelet extractions, which produced knowledge during our previous study [8] of the excellent some interfering artifacts that could compromise the ability capability of SWT in extracting these specific frequency to review the scalp topographies by the medical experts and subbands. mislead them in diagnosing the cause of these parasites. )e GOF results showed a similar pattern across the Figure 9 displays the scalp topographies of the original signal different frequency subbands and different SNR values with compared to both the mother wavelet extraction and the a distinct superiority to sym20, coif5, bior6.8, rbio6.8, and contaminated scalp topographies of Haar and sym4. As an dmey wavelets. assessment of the PSD and scalp topography evaluation, the In order to explore and investigate this superiority, we sym20 and demy mother wavelets demonstrated the best have extracted the best mother wavelets of every wavelet results while the Haar and sym4 produced the worst ones. family and the wavelets that already showed some note- worthy results in other studies, such as sym4 in [8], db5, and sym6 in [10] and sym9 in [9, 11], in every EEG frequency 3.3. /e Source Localization. For the source localization, we subband, as shown in Figure 6. performed the Independent Component Analysis (ICA) on Besides, after isolating the GOF results about the limited the extracted signals by the noteworthy mother wavelets; number of noteworthy wavelets, we notice that the perfor- then, we used the BEM for the forward problem and ECD for mance of the wavelet extraction changes from one frequency the inverse problem. As we have already mentioned, the ICA subband to another with an obvious preeminence in gamma is a computationally costly process for feature extraction, and alpha waves. We also observe that sym4 in [8] is the lowest especially with 62 EEG channels for the extraction of the in the GOF performance due to the approximated coefficient same number of components before the source localization, choice in the decomposition phase compared to our choice of so we reduced the process to include only the alpha and approximated coefficient in this study for all the wavelets. gamma frequency subbands. )e alpha wave is the most Finally, to lock the GOF evaluation results, we calculated important brainwave activity in the human brain and the the noteworthy wavelet average across the five frequency gamma wave is perceived as an indicator of high active subbands and ordered them from the lowest performance, cognitive state and constantly used in brain malfunction and on the left, to the best performance, on the right by their disease confirmation [26]. GOF score in Figure 7. In fact, the best results were achieved )e ICA was performed using the runica algorithm from using demy and sym20 wavelets, while the worst results used the EEGLAB toolbox [27]. )en, the BEM and ECD were sym4 [8] and Haar also. executed using the fieldtrip toolbox [28]. We set a rejection threshold for the components based on the Residue Variance equivalent to RV � 15% as it is the optimum value in 3.2. /e PSD and Topographies Evaluation Results. )e component rejection, as confirmed by Artoni et al. [29]. In Power Spectral Density (PSD) is also an important Sym4 [5] Sym1/db1 (haar) Sym2/db2 Sym3/db3 Sym4/db4 Sym5/db5 Sym6/db6 Sym9/db9 Sym20/db20 Coif1 Coif2 Coif3 Coif4 Coif5 Bior1.1 Bior1.3/bior2.2/bior3.1 Bior1.5/bior2.4/bior3.3 Bior2.6/bior3.5/bior4.4 Bior2.8/bior3.7/bior5.5 Bior3.9 Bior6.8 Rbio1.1 Rbio1.3/rbio2.2/rbio3.1 Rbio1.5/rbio2.4/rbio3.3 Rbio2.6/rbio3.5/rbio4.4 Rbio2.8/rbio3.7/rbio5.5 Rbio3.9 Rbio6.8 Dmey (meyer) 6 Journal of Healthcare Engineering The average of GOF (across the SNR values of –5, 10, and 15) for every extracted frequency subband Sym4 [5] Haar Sym4 Db5 Sym6 Bior6.8 Sym9 Coif5 Sym20 Demy Rbio6.8 Delta Beta Theta Gamma Alpha Figure 6: )e results of the average GOF for every EEG frequency subband with the selected mother wavelets across the SNR values of −5, 10, and 15 dB. GOF total average across the 5 EEG frequency sub-bands 91.01 90.6 89.56 89.85 88.7 89.07 88.26 90 86.89 85.79 66.56 62.46 Sym4 [5] Haar Sym4 Db5 Bior6.8 Rbio6.8 Sym9 Coif5 Sym6 Sym20 Demy Figure 7: )e GOF average of the noteworthy wavelets across the EEG frequency subbands and SNR values ranked from left to right by order of best performance. the order of a component, the more data (neural and/or Figure 10, we present the source localization of the alpha and gamma extracted waves using the different noteworthy artifactual) it accounts for [30]. Figure 11 shows the number of components localized by mother wavelets. As we can see, every mother wavelet ex- traction has a different number of sources localized under each mother wavelet in the alpha and gamma frequency the Residue Variance (RV) error threshold and different subbands and only the number of components that were not source locations compared to each other. In order to localized in the other frequency subbands. evaluate the source localization of our different mother An interpretation of the number of localized component wavelets, we focus on the number of localized components results showed that the sym20 mother wavelets produced the by every mother wavelet and the number of times every best results followed by Haar and bior6.8, while coif5 had the mother wavelet has the best accuracy (lower RV value) in lowest number of localized components. localizing the source of a component and the average of In Figure 12, we explore the accuracy of the noteworthy mother wavelets in source localization by comparing the accuracy in the five first components. )e reason for which we have included the accuracy of the five first components in number of times each mother wavelet managed to record the our evaluation is that the ICA using the runica algorithm for lowest RV score. )is chart also considers the localization in the output components in a decreasing order of the EEG the alpha wave, gamma wave, and the combined best-lo- variance accounted for by each component, that is, the lower calized components of both frequency subbands. Journal of Healthcare Engineering 7 Original EEG signal –5 510 15 20 25 Frequency 20 20 15 15 10 10 Haar Sym4 5 5 0 0 –5 –5 510 15 20 25 510 15 20 25 Frequency Frequency 20 20 15 15 10 10 db5 Sym6 5 5 0 0 –5 –5 510 15 20 25 510 15 20 25 Frequency Frequency 20 20 15 15 10 10 Bior 6.8 Sym9 5 5 0 0 –5 –5 510 15 20 25 510 15 20 25 Frequency Frequency 20 20 15 15 10 10 Rbio 6.8 Coif5 5 5 0 0 –5 –5 510 15 20 25 510 15 20 25 Frequency Frequency 20 20 15 15 10 10 Dmey Sym20 5 5 0 0 –5 –5 510 15 20 25 510 15 20 25 Frequency Frequency Figure 8: PSD visualization of the different noteworthy mother wavelets in alpha wave extraction. 3.0 6.0 10.0 20.0 45.0 Original dataset Detected changes Haar Sym4 Other mother wavelets Delta Theta Alpha Beta Gamma Figure 9: A scalp topographies comparison between the original dataset, the noteworthy mother wavelets extractions, and the contaminated scalp topographies of Haar and sym4 wavelets for the five EEG frequencies subbands. )e sym20 mother wavelet scored the best accuracy original EEG signal had an impressive accuracy in gamma results followed by Haar and sym9, while rbio 6.8 did not wave, which indicates the interference of the other frequency have even once the best accuracy compared to the other subbands or the 50 Hz noise artifact and compromised the wavelets for both frequency subbands. We also spot that the integrity of the located sources considering that the gamma Log power spectral Log power spectral Log power spectral Log power spectral Log power spectral ∗ 2 ∗ 2 ∗ 2 ∗ 2 ∗ 2 density 10 log (µV /Hz) density 10 log (µV /Hz) density 10 log (µV /Hz) density 10 log (µV /Hz) density 10 log (µV /Hz) 10 10 10 10 10 Log power spectral ∗ 2 density 10 log (µV /Hz) Log power spectral Log power spectral Log power spectral Log power spectral Log power spectral ∗ 2 ∗ 2 ∗ 2 ∗ 2 ∗ 2 density 10 log (µV /Hz) density 10 log (µV /Hz) density 10 log (µV /Hz) density 10 log (µV /Hz) density 10 log (µV /Hz) 10 10 10 10 10 8 Journal of Healthcare Engineering Original dataset Alpha Gamma Alpha Gamma Haar db5 Coif5 Sym4 Sym6 Bior 6.8 Sym9 Rbio 6.8 Sym20 dmey Figure 10: Visualization of the alpha and gamma waves source localization using the noteworthy mother wavelets extractions. Source localization components of alpha and gamma with RV error under 15% 14 14 13 13 15 25 14 8 8 15 6 8 88 6 8 12 12 12 12 11 11 11 11 11 Original Sym4 Sym6 Sym9 Sym20 Haar Db5 Coif5 Bior6.8 Rbior6.8 Demy dataset Alpha components Gamma components Unique components in alpha & gamma Figure 11: )e number of components localized by the noteworthy mother wavelets in alpha and gamma waves with RV under 15%. Journal of Healthcare Engineering 9 Poll of the best component RV error minimization score Original Sym4 Sym6 Sym9 Sym20 Haar Db5 Coif5 Bior6.8 Rbior6.8 Demy dataset Alpha component localization Gamma component localization Both alpha & gamma component localization Figure 12: )e number of times each mother wavelet scored the best accuracy for source localization in alpha and gamma waves. Table 2: )e RV average of 5 first components for the noteworthy mother wavelets. Alpha RV average of 5 first Gamma RV average of 5 first Combined RV average of best 5 first Wavelet components components components Original 10.8 dataset Sym4 8.9 9.5 8.6 Sym6 7.8 9.6 7.7 Sym9 6.8 10.2 6.8 Sym20 9.4 9.6 6.2 Haar 10.9 8.3 6.7 Db5 7.8 9.7 7.8 Coif5 8.8 10.2 6.2 Bior6.8 6.7 9.6 6.6 Rbio6.8 9.3 10.4 9.3 demy 9.4 9.6 6.8 wave had poor frequency energy that could not produce As an overall perception of the source localization results such result. in evaluating our mother wavelets, we can classify the sym20 Table 2 presents the final criteria for source localization mother wavelet as the best mother wavelet extraction overall, while the Haar occupies the second place with questionable evaluation that focuses on the accuracy of the five first components. )e accuracy is expressed with the RV values, results due to our previous readings of the GOF, PSD, and which means that the lower the RV value is, the better scalp topographies that proved the interference of frequency accuracy will be. overlapping and noise artifacts in the sincerity of the lo- Actually, the best result for the alpha wave was achieved calized components. If we eliminate the Haar mother by bior6.8 while the worst was recorded by the Haar mother wavelet, we must crown the bior6.8 mother wavelet the wavelet. For the gamma wave, the Haar mother wavelet second place considering the number of localized compo- produced the best result, while rbio6.8 extractions were last nents and the best results achieved in the accuracy average of compared to the other wavelet extractions. )en, regarding the first components in the alpha wave followed by the coif5 the combined best-localized components of alpha and and sym9 mother wavelets. While On the other hand, the gamma, the sym20 and coif5 shared the first place in dmey produced a somehow moderate result in light of the extracting the most accurate first five components, with the promising potential in the earlier evaluations of GOF, PSD, rbio6.8 mother wavelet in the last place. and scalp topographies. )e least favorite mother wavelet in 10 Journal of Healthcare Engineering [3] L. Murali, D. Chitra, T. Manigandan, and B. Sharanya, “An source localization was the rbio6.8 with the worst accuracy efficient adaptive filter architecture for improving the seizure results recorded by all the noteworthy mother wavelets. detection in EEG signal,” Circuits, Systems, and Signal Pro- cessing, vol. 35, no. 8, pp. 2914–2931, 2016. 4. Conclusion [4] V. Singh, K. Veer, R. Sharma, and S. Kumar, “Comparative study of FIR and IIR filters for the removal of 50 Hz noise In this paper, we have compared 51 different mother from EEG signal,” International Journal of Biomedical Engi- wavelets taken from 7 different families including Haar, neering and Technology, vol. 22, no. 3, pp. 250–257, 2016. Symlets, Daubechies, Coiflets, Discrete Meyer, Biorthogonal, [5] S. S. Nallamothu, R. K. Dodda, and K. S. Dasara, Eye Blink and reverse Biorthogonal, which are applied to source Artefact Cancellation in EEG Signal Using Sign-Based Nonlinear localization and extraction of EEG signal. For the source Adaptive Filtering Techniques. Information Systems Design localization performance comparison, the 10 mother wavelets and Intelligent Applications, Springer, Singapore, Singapore, selected from the 51 mother wavelets produced an adequate result. However, the sym20 outshined all the other wavelets [6] K. D. Tzimourta, Epileptic Seizures Classification Based on Long-Term EEG Signal Wavelet Analysis. Precision Medicine and took the lead almost in every evaluation followed by a Powered by Health and Connected Health, Springer, Singa- notable performance from bior6.8, coif5, and sym9, re- pore, Singapore, 2018. spectively. )en, the least results were produced by the Haar [7] C. Kandilli and B. Mertoglu, “Optimisation design and and rbio6.8 mother wavelets. As a conclusion, the Symlet operation parameters of a photovoltaic thermal system family generates the top results for EEG signal, as dem- integrated with natural zeolite,” International Journal of onstrated by our study. )en, bior6.8 and coif5 are the Hydromechatronics, vol. 3, no. 2, pp. 128–139, 2020. second important mother wavelets for source localization. [8] N. Abdennour, “Extraction and localization of non-con- Regarding the evaluation methods, we used the goodness taminated alpha and gamma oscillations from EEG signal of fit (GOF), the Power Spectral Density, and scalp to- using finite impulse response, stationary wavelet trans- pographies in the extraction of EEG frequency subbands form, and custom FIR,” in Proceedings of the International applied to benchmarks containing source localization with Conference on Artificial Neural Networks, Springer, Cham, the number of located sources and accuracy of localization. Germany, July 2018. )e source localization is produced via Stationary Wavelet [9] A. H. Akkar and F. A. Jasim, “Optimal mother wavelet Transform (SWT) and an Independent Component Analysis function for EEG signal analyze based on packet wavelet (ICA) feature extraction followed by Boundary Element transform,” International Journal of Scientific & Engineering Model (BEM) and Equivalent Current Dipole (ECD) so- Research, vol. 8, pp. 1222–1227, 2017. [10] L. Condo and L. Efren, “Comparison of wavelet transform lutions for the forward and inverse problem. Future studies symlets (2–10) and daubechies (2–10) for an electroenceph- and advancements could explore the improvement of the alographic signal analysis,” in Proceedings of the 2017 IEEE source localization feature extraction or forward and inverse XXIV International Conference on Electronics, Electrical En- problem solutions. )e use of artificial intelligence tech- gineering and Computing (INTERCON), IEEE, Cagliari, Italy, niques based on the deep neural network could help to August 2017. facilitate the simulation and give better results. [11] A.-Q. Noor, “Selection of mother wavelet functions for multi- channel EEG signal analysis during a working memory task,” Data Availability Sensors, vol. 15, no. 11, pp. 29015–29035, 2015. [12] K. Trimeche, Generalized Wavelets and Hypergroups, Rout- )e medical used data are available from the corresponding ledge, London, UK, 2019. author upon request. [13] M. Safa, M. Ahmadi, J. Mehrmashadi et al., “Sedghi Selection of the most influential parameters on vectorial crystal growth of highly oriented vertically aligned carbon nanotubes by Conflicts of Interest adaptive neuro-fuzzy technique,” International Journal of Hydromechatronics, vol. 3, no. 3, pp. 238–251, 2020. )e authors declare no conflicts of interest. [14] M. Rhif, A. Ben Abbes, I. Farah, B. Mart´ınez, and Y. Sang, “Wavelet transform application for/in non-stationary time- Acknowledgments series analysis: a review,” Applied Sciences, vol. 9, no. 7, pp. 1345, 2019. Dr. Omar Cheikhrouhou thanks Taif University for its [15] S.-H. Wang, Y.-D. Zhang, Z. Dong, and P. Phillips, Wavelet support under the project Taif University Researchers Families and Variants. Pathological Brain Detection, Springer, Supporting Project no. TURSP-2020/55, Taif University, Singapore, Singapore, 2018. Taif, Saudi Arabia. [16] C. Sedghi, W. Yan, X. Cai, S. Liu, T. H. Li, and G. Li, “Neural saliency algorithm guide bi-directional visual perception style transfer,” CAAI Transactions on Intelligence Technology, vol. 5, References no. 1, pp. 1–8, 2020. [17] X. Zhang, “Localizing percentages of interictal 18 F-fluo- [1] H. Berger, “Uber das elektrenkephalogramm des menschen,” rodeoxyglucose (FDG)-PET and magnetoencephalography Archiv fur ¨ Psychiatrie und Nervenkrankheiten, vol. 87, no. 1, pp. 527–570, 1929. (MEG) in pre-surgical evaluation of 107 consecutive patients with non-lesional epilepsy,” International Journal of Clinical [2] S. Sanei, Adaptive Processing of Brain Signals, John Wiley & Sons, Hoboken, NJ, USA, 2013. & Experimental Medicine, vol. 9, p. 12, 2016. Journal of Healthcare Engineering 11 [18] O. A. Petroff, D. D. Spencer, I. I. Goncharova, and H. P. Zaveri, “A comparison of the power spectral density of scalp EEG and subjacent electrocorticograms,” Clinical Neurophysiology, vol. 127, no. 2, pp. 1108–1112, 2016. [19] N. Mammone, G. Inuso, F. La Foresta, M. Versaci, and F. C. Morabito, “Clustering of entropy topography in epileptic electroencephalography,” Neural Computing and Applica- tions, vol. 20, no. 6, pp. 825–833, 2011. [20] J. Song, C. Davey, C. Poulsen et al., “EEG source localization: sensor density and head surface coverage,” Journal of Neu- roscience Methods, vol. 256, pp. 9–21, 2015. [21] H. Azizollahi, A. Aarabi, and F. Wallois, “Effects of uncer- tainty in head tissue conductivity and complexity on EEG forward modeling in neonates,” Human Brain Mapping, vol. 37, no. 10, pp. 3604–3622, 2016. [22] L. Gaul, M. Kogl, ¨ and M. Wagner, Boundary Element Methods for Engineers and Scientists: An Introductory Course with Advanced Topics, Springer Science & Business Media, Berlin, Heidelberg, Germany, 2013. [23] N. S. Elaina, “Somatosensory source localization for the magnetoencephalography (MEG) inverse problem in patients with brain tumor biomedical Engineering (ICoBE),” in Pro- ceedings of the 2015 2nd International Conference on IEEE, New Delhi, India, March 2015. [24] K.-L. Du and M. N. S. Swamy, Independent Component Analysis. Neural Networks and Statistical Learning, Springer, London, UK, 2014. [25] Z. Ali and T. Mahmood, “Complex neutrosophic generalised dice similarity measures and their application to decision making,” CAAI Transactions on Intelligence Technology, vol. 5, no. 2, pp. 78–87, 2020. [26] N. Hiroki, “Scalp EEG ictal gamma and beta activity during infantile spasms: evidence of focality,” Epilepsia, vol. 58, no. 5, pp. 882–892, 2017. [27] A. Delorme and S. Makeig, “EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including inde- pendent component analysis,” Journal of Neuroscience Methods, vol. 134, no. 1, pp. 9–21, 2004. [28] R. Oostenveld, P. Fries, E. Maris, and J.-M. Schoffelen, “FieldTrip: open source software for advanced analysis of MEG, EEG, and invasive electrophysiological data,” Com- putational Intelligence and Neuroscience, vol. 2011, Article ID 156869, 9 pages, 2011. [29] F. Artoni, D. Menicucci, A. Delorme, S. Makeig, and S. Micera, “RELICA: a method for estimating the reliability of independent components,” NeuroImage, vol. 103, pp. 391– 400, 2014. [30] A. Delorme, T. Sejnowski, and S. Makeig, “Enhanced de- tection of artifacts in EEG data using higher-order statistics and independent component analysis,” Neuroimage, vol. 34, no. 4, pp. 1443–1449, 2007. [31] M. Kaur and D. Singh, “Multi-modality medical image fusion technique using multi-objective differential evolution based deep neural networks,” Journal of Ambient Intelligence and Humanized Computing, vol. 12, no. 2, pp. 2483–2493, 2021. [32] M. Kaur and D. Singh, “Fusion of medical images using deep belief networks,” Cluster Computing, vol. 23, no. 2, pp. 1439–1453, 2020. [33] B. R. Murlidhar, R. K. Sinha, E. T. Mohamad, R. Sonkar, and M. Khorami, “)e effects of particle swarm optimisation and genetic algorithm on ANN results in predicting pile bearing capacity,” International Journal of Hydromechatronics, vol. 3, no. 1, pp. 69–87, 2020.
Journal
Journal of Healthcare Engineering – Hindawi Publishing Corporation
Published: Apr 28, 2021