Approximation-Aided Epilepsy Detection Using Linear and Non-Linear Classifiers

Introduction

Out of the several disorders of the human brain, Epilepsy seems to be prominent and most occurring, Identification of Epilepsy patients is critical for the quick treatment and recovery. Epilepsy could be identified by analyzing the EEG signals of the human. In (Güzel, 2019), the algorithm hybridized the linear and nonlinear classifiers with the discrete wavelet transforms (DWT) and principal component analysis (PCA), separately. The model estimation process with multivariate logistic regression (MLR) and artificial neural networks (ANNs) was used to control the complexity of model and reduce the dimension of feature matrix, and implementation was done using PCA. In addition, to quantify the complexity and select the best models, the information criteria were considered for both MLR and ANNs. To improve the classification performance, ANNs were trained by various gradient algorithms and early stopping and cross-validation techniques were used. In (Martinez-del-Rincon et al., 2017), a novel approach was proposed for automatic epilepsy seizure detection based on EEG analysis that exploited the underlying non-linear nature of EEG data. The research involved two main contributions namely the use of non-linear classifiers through the so-called kernel trick and the proposal of a Bag-of-Words model for extracting a non-linear feature representation of the input data in an unsupervised manner. The obtained results proved the robustness of the proposed solution to more realistic for variable conditions.

In (Siddiqui et al., 2020), detailed review of various approaches to seizure detection using machine learning classifiers and statistical features, was done. The main challenges were selection of appropriate classifiers and features. An overview of the wide varieties of these techniques over the last few years based on the taxonomy of statistical features and machine learning classifiers ‘black-box’ and ‘non-black-box’, was done. Globally, epilepsy affects approximately 50 million people, with 100 million being affected at least once in their lifetime (WHO, 2022). Applications of machine learning are significantly seen on health and biological data sets for better outcomes (Mahmud et al., 2020). Researchers/scientists on different areas, specifically, data mining and machine learning, are actively involved in proposing solutions for better seizure detection. Machine learning plays a significant and potential role in solving the problems of various disciplines like healthcare (Singh & Gupta, 2019). Seizure detection has been done by various machine learning classifiers such as ANN, SVM, decision tree, decision forest, and random forest (Abbasi & Goldenholz, 2019).

Javad Birjandtalab et al. (Andrzejak et al., 2001) used ANN with a weighted cost function to imbalanced EEG dataset, by achieving 86% F-measure. In 2018, Lahmiri et al. (Birjandtalab et al., 2018) used generalized Hurst exponent (GHE) and KNN, to propose a system for identifying the ‘seizure’ and ‘non-seizure’ classes from intracranial EEG recordings, detection rate, with 100% accuracy rate. In another study by Al Ghayab et al. (Lahmiri & Shmuel, 2018), the concept of Information gain theory was used, to extract and rank the meaningful features from EEG signal dataset. The least square-support vector machine (LS-SVM) was then applied to classify the seizure cases. Rand and Sriram (Al Ghayab et al., 2019) used four classifiers such as SVM, KNN, random forest, and Adaboost on a high-dimensional dataset prepared by 28 features. Their result shows that SVM outperformed on the cubic kernel. In another study, Manzouri et al. (Raghu & Sriraam, 2018) used SVM and random forest on the dataset produced by 10-time and frequency features. In comparison to SVM-based detector, random forest classifier outperformed by a huge margin. Subasi et al. (Manzouri et al., 2018) achieved 100% of accuracy using four machine learning classifiers such as ANN, KNN, SVM, and random forest on two popular datasets—Freiburg and CHB-MIT to classify the three different states of seizures ‘pre-ictal’, ‘ictal’, and ‘inter-ictal.