Introduction
Technological innovation has improved various processes including healthcare intervention systems. Epilepsy is no exception as medical experts have developed automatic detection machines for analysis of electroencephalogram (EEG) to sense electrical processes in the brain using electrodes to instantiate seizure. Programmed seizure recognition is essential for closed-loop receptive cortical activation processes. This review evaluates various literature on recent technology on epileptic extraction, seizure detection, and prediction strategies. Machine learning techniques are quick and accurate in epileptic seizure detection.
Epileptic Extraction Methods
The two key phases of developing an automatic epilepsy detection system are feature selection and classification. The classifier issues a proper class code to the derived function vector after feature extraction decreases the proportions of the input indicator by preserving descriptive features. In a recent study, Niknazar et al. proposed successful feature extraction strategies for automatic epileptic EEG wave recognition. To distinguish epileptic EEG signals in the sample, two successful feature extraction strategies were introduced. Various machine learning models were used to differentiate epileptic seizure and non-seizure signs. The standard epilepsy EEG database given by the University of Bonn was applied in the study.
The detection accuracy was assessed by a 10-fold cross authentication procedure. Decision Tree (DT), Support Vector Machine (SVM), Artificial Neural Network (ANN), and Nearest Neighbor (NN) classifiers are included. The tests were replicated 50 times and for the detection of natural and epileptic EEG signs, 1D-LGP and LNDP function extraction strategies with ANN Classification algorithm achieved average precision of 99.82 percent and 99.80 percent, respectively. The classification findings outperformed several current approaches as 1D-LGP and LNDP were useful extraction strategies for classifying epileptic EEG activities.
Automatic epileptic seizure detection can be efficient in preventing any related issues. In a recent research, Niknazar et al., evaluated EEG-based technology proposed for early detection of seizure to identify the most efficient ones for sensing seizure. The authors extracted seizure tracking properties from intracranial EEG waves obtained through intrusive pre-surgical epilepsy screening of patients with drug-resistant focal seizures at a University Hospital in Freiburg’s. Time, intensity, and process space domain features, and similarity/dissimilarity attributes, were factored. The quantitative research approach, ANOVA, was used to examine the output of each function. Various performance analyses were completed on reports from networks in the seizure-onset region and measurements from channels beyond the seizure-onset area. Correlation elements that quantify variable characteristics of the EEG indicator and the varying dynamics of seizures could distinguish ictal states from pre-ictal conditions (p < 0.01).
Of these, the authors’ proposed function, Bhattacharyya-based dissimilarity indicator, passed a post-hoc examination, indicating that it could differentiate pre-ictal and post-ictal cycles from ictal phases. Bhattacharyya-based dissimilarity sensor was then used to track epileptic seizures, indicating no major change in function output between SOZ-in and SOZ-out reports. Statistical analyses were used to assess the discriminative effect of EEG seizure identification features. As a consequence, the best features to choose for an accurate seizure monitoring device optimized for patients with drug-resistant temporal lobe epilepsy, was the similarity/dissimilarity scales.
Epileptic Seizure Detection Method
Seizures may have an adverse effect on patients mental, social, and physical life. Thus, their diagnoses are largely reliant on laborious manual curation by skilled physicians using EEG signals. The majority of current EEG-based seizure detectors are patient-dependent, requiring a trained detection algorithm for each user. Hence, a new patient can only use it after numerous episodes of the seizures, making it ineffective. In a study, Yang et al. investigated patient independent sensor of epileptic events using CHB-MIT Scalp EEG. To assess the topological trends of the EEG activities, an innovative function extraction technique known as MinMaxHist is suggested.
The EEG detectors are then parameterized using MinMaxHist and other feature extraction techniques. Later, a systematic set of function scanning and identification optimization tests are performed, and eventually, an improved EEG-based seizure classification method of 30 functions is introduced, with overall values for precision, Matthew’s correlation coefficient, tolerance, accuracy, and Kappa. The system with MinMaxHist technologies had a 0.0464 higher classification than the model excluding MinMaxHist functions. The suggested technique outperformed existing approaches in precision and efficiency.
Seizure detection from EEG activity can aid neurologists in analyzing the statuses of epileptic patients. The diversity of epileptic seizures makes it difficult to distinguish the sequence of epilepsy signs from natural ones. Wulandari et al. address the characterization of seizure and non-seizure disorders of epilepsy centered on EEG signal spectrum characteristics. Empirical Mode Decomposition was applied in extracting these functions. These characteristics were loaded into the Support Vector Machine as data. The authors suggested combining the first 4 IMFs to derive frequency functions, which they tested on two data sets containing only waves from extracranial EEG. The findings match those of Ynag et al. indicating that for SVM kernel activities, the metrics of precision and accuracy using the multiple features of the first selected IMFs outpaced those using single IMF features.
A technology that could alert all patients and physicians to the imminent epilepsy occurrence would vastly improve patients’ life. Deriche et al. performed research on epilepsy detection suggesting that time and frequency (TF) analysis be used to extract factors capable of distinguishing between regular and abnormal EEG residues. The parameters were derived from the EEG detector Time Frequency vector through Singular Value Decomposition. The results reveal that, regardless of the classification model used, most conventional classification strategies yielded excellent seizure identification results when combined with the suggested TF functionality. The results are consistent with those of Wulandari et al., who found that the novel innovative technology features improve the seizure detection techniques.
Accurate epileptic seizure detection improves a patient’s response to attacks and overall quality of life. Dash et al. conducted an iterative filtering breakdown EEG alerts to improve seizure detection accuracy. The authors evaluate the suggested approach using Indian digital databases. They apply the iterative filtering decay approach to derive sub-elements from the EEG signal. The approach attains an accuracy of over 99% in seizure detection. The findings clearly match those of Wundari et al. and Deriche et al. that innovative seizure detection techniques are more accurate in detecting epilepsy. Acharya et. use the convolutional neural network to assess EEG signals. In particular, they employed a 13-layer convolutional neural system on dataset from five patients. The technique attained an average accuracy of about 90%, which is a good score. The findings match those of the previous researchers Wundari et al., Deriche et al., and Dash et al. that novel epileptic seizure detectors are accurate.
More researchers investigate the effectiveness of automated detection approaches in assessing seizures. For instance, Sudalaimani et al. also proved the accuracy of innovative seizure detection approached through the use of sub-frequency EEG data from a previous signal. Sharma et al. found an over 78% accuracy in a proposed system; a wavelet decomposition. Zazzaro et al. found a 99% accuracy of a trained classifier through indicator processing matching the findings of Sudalaimani et al. and Sharma et al.. Ullah et al. tested two augmentation systems on a university dataset whose findings confirmed the suitability of the approach in detecting epilepsy. Lastly, Nkengfack et al. also tested a full dispensation network of analysis for seizures. The findings supported the efficiency of the suggested processing chain with an accuracy of 96.25 to 100 percent. Overall, the studies provide matching findings on the efficiency of novel proposed seizure detection systems on recognizing epileptic attacks.
Epileptic Seizure Prediction Methods
Various authors also investigate the role of machine learning method in predicting seizures. Savadkoohi et al. used EEG signal to assess brain electrical activities. They examined the best approach to identify meaningful characteristics from an epileptic EEG. The findings revealed that SVM had a minimal edge over KNN. In another study, Wei et al. converted EEG into two-dimensional figures for multi-network fusion. A sustainable recurring network was suggested to offer a spatiotemporal extensive learning framework to detect seizures. The findings match those of Savadkoohi et al. that the novel prediction model is accurate with a 93.4% precision. Tsiouris et al. used Long-Short Term Memory (SLTM) as that of Wei et al. The findings showed increased seizure prediction using the SLTM as compared to conventional approaches replicating the findings of Wei et al. San-segundo et al. also evaluated various EEG signals using public databases. The findings indicated improved accuracy in detecting seizure. Overall, the reviews attest the importance of machine learning methods in predicting seizures.
Conclusion
Machine learning methods are efficient in early and accurate seizure detection. In this study, various empirical researches are evaluated to understand the application and accuracy of the strategy. Most researches prove that indeed, automated machine detection approaches are efficient in enhancing epileptic seizure detection to improve the patients’ quality of life. In particular, the proposed technologies include EEG, Long-Short Term Memory, KNN, and other automated detection systems. The findings support machine learning in quick seizure recognition to avoid associated dangers.
Reference List
H. Niknazar., S.R. Mousavi., M. Niknazar., V. Mardanlou., B. N. Coelho. “Performance analysis of EEG seizure, detection features.” Epilepsy Research, vol. 167, pp. 1-8, 2020. Web.
A. B. Jaiswal, and H. Banka. “Local pattern transformation-based feature extraction techniques for classification of epileptic EEG signals.” Biomedical Signal Processing and Control vol. 34, pp. 81-92. 2017. Web.
S. Yang., B. Li., Y. Zhang., M. Duan., S. Liu., Y. Zhang/. X. Feng., R. Tan., L. Huang., F. Zhou. “Selection of features for patient-independent detection of seizure events using scalp EEG signals.” Computers in biology and medicine vol. 119, pp. 1-13, 2020. Web.
D. P. Wulandari, N. G. P. Putri, Y. K. Suprapto, S. W. Purnami, A. I. Juniani, W. R. Islamiyah. “Epileptic seizure detection based on bandwidth features of EEG signals.” Procedia Computer Science vol. 161, pp. 568-576, 2019. Web.
M. Deriche, S. Arafat, S. Al-Insaif, M. Siddiqui. “Eigenspace time frequency-based features for accurate seizure detection from EEG data.” IRBM vol. 40, no. 2, pp.122-132, 2019. Web.
D. P. Dash, M. H. Kolekar, K. Jha. “Multi-channel EEG based automatic epileptic seizure detection using iterative filtering decomposition and Hidden Markov Model.” Computers in biology and medicine vol. 116, pp. 1-11, 2020. Web.
U. R. Acharya, S. L. Oh, Y, Hagiwara, J. H, Tan, H, Adeli. “Deep convolutional neural network for the automated detection and diagnosis of seizure using EEG signals.” Computers in biology and medicine vol. 100, pp. 270-278, 2018. Web.
C. Sudalaimani, N. Sivakumaran, T. T. Elizabeth, V. S. Rominus. “Automated detection of the preseizure state in EEG signal using neural networks.”biocybernetics and Biomedical engineering vol. 39, no. 1, pp. 160-175, 2019. Web.
M. Sharma, S. Patel, and U. R. Acharya. “Automated detection of abnormal EEG signals using localized wavelet filter banks.” Pattern Recognition Letters vol. 133, pp. 188-194, 2020. Web.
G. Zazzaro, S. Cuomo, A. Martone. “Eeg signal analysis for epileptic seizures detection by applying data mining techniques.” Internet of Things. 2019. Web.
I. Ullah, M. Hussain, E. Qazi. “An automated system for epilepsy detection using EEG brain signals based on deep learning approach.”Expert Systems with Applications, vol.107, pp. 61.71, 2018. Web.
L. C. D. Nkengfack, D. Tchiotsop, R. Atangana, V. Louis-Door, D. Wolf. “EEG signals analysis for epileptic seizures detection using polynomial transforms, linear discriminant analysis and support vector machines.” Biomedical Signal Processing and Control vol. 62, pp. 1-14, 2020. Web.
M. Savadkoohi, T. Oladunni, L. Thompson. “A machine learning approach to epileptic seizure prediction using Electroencephalogram (EEG) Signal.” Biocybernetics and Biomedical Engineering vol. 40.no. 3, 1328-13411, 2020. Web.
X. Wei, L. Zhou, Z. Zhang, Z. Chen, Y. Zhou. “Early prediction of epileptic seizures using a long-term recurrent convolutional network.” Journal of neuroscience methods vol. 327, pp. 1-10, 2019. Web.
K. M. Tsiouris, V. C. Pezoulas, M. Zervakis, S. Konitsiotis, D. D. Koutsouris, D. I. Fotiadis. “A long short-term memory deep learning network for the prediction of epileptic seizures using EEG signals.” Vol. 99, 24-37, 2018. Web.