Adaptive rag-bull rider: A modified self-adaptive optimization algorithm for epileptic seizure detection with deep stacked autoencoder using electroencephalogram

Highlights

• Proposes an Adaptive rag-Rider optimization algorithm (Adaptive rag-ROA)-based Deep SAE to detect seizures from EEG.

• Here, Adaptive rag-ROA is used to train deep-stacked autoencoder (Deep SAE) for discovering epileptic seizures.

• Adaptive rag-ROA is devised by incorporating the Adaptive concept in rag-ROA.

• The analysis is carried out on TUH EEG Epilepsy Corpus and CHB-MIT Scalp EEG Database.

• The proposed Adaptive rag-ROA-based Deep SAE has the accuracy of 91.5%, sensitivity of 85.2%, and specificity of 86%.

Abstract

Electroencephalogram (EEG) signal is widely adapted for monitoring epilepsy to rejuvenate the close-loop brain. Various conventional techniques are devised for identifying seizures that depend on visually analyzing EEG signals that are an expensive and complicated process if there is a rise in numbers of the channel. A new technique, namely Adaptive rag-Rider optimization algorithm (Adaptive rag-ROA) is presented to train deep-stacked autoencoder (Deep SAE) for discovering epileptic seizures. Initially, the EEG signals are given to the pre-processing module, in which noise has been removed by the bandpass filter. Then, the noise removed signals are provided as an input wherein the EEG is divided into various channels and each channel performs the extraction of features. Here, the features such as Taylor-based delta AMS, Holoentropy, fluctuation index, relative energy, tonal power ratio, spectral features, and linear prediction coefficient (LPC) are acquired from each channel. Furthermore, the Probabilistic principal component analysis (PPCA) is adapted to diminish the dimensionality of features. The obtained feature vectors are fed to Deep SAE for epileptic seizure recognition. The Deep SAE training is carried out with Adaptive rag-ROA that is devised by incorporating the Adaptive concept in rag-ROA. Thus, the output generated from the proposed Adaptive rag-ROA-based Deep SAE is adapted to detect seizures from EEG. The proposed Adaptive rag-ROA-based deep SAE outperformed revealing the highest accuracy of 91.5%, the sensitivity of 85.2%, and specificity of 86%.

Introduction

The infirmity of the nerve occurs because of the electrical expulsion of cortical neurons contained in the brain and is commonly known as Epilepsy that is susceptible to produce various types of seizures. These types of seizures are unexpected, unforeseen, and unprovoked because of its instant factors. There exist individuals over 65 million who experience these kinds of disorders. There exists a 75% seizures case that is diagnosed using therapies [9]. With residual 25% cases, the seizures ruin despite drugs, and it impudent patients surviving with seizures [10]. The patients suffering from epilepsy is typically inaccessible overnight and prone to different corporeal pain or suffocation that are caused because of the barren airway. There exists a need for support with small delays after the initiation of seizures that urge rupture. The seizure is generally risky at nighttime while patients are alienated and could not call for assistance. There exist some nighttime seizures that are not spotted by patients and may cause several therapeutic obstructions or may lead to death. There exist requirements to discover seizures in real-time that can promote notice to the individuals who reside close whenever the seizure is determined. The stipulation of appropriate assistance can reduce the rate of mortality and prevent intricacy [6].

Epilepsy is diagnosed medically by performing different evaluations considering positron emission tomography (PET), magneto-encephalogram (MEG), computed tomography (CT), EEG, and magnetic resonance imaging (MRI). The EEG is termed as the best modality in contrast to other techniques because of high resolution in temporal data and cost-effective property. The stipulation of undeviating capacity is achievable by EEG because of certain actions of the brain. The EEG is an effective modality that is used to observe and detect epileptic seizures that regularly produce divergence in evaluated EEG. Several techniques are stated to discover seizures with EEG [13,14,6]. On the other hand, the analysis of the patient’s EEG considering different days is required for accurate assessment and seizure identification. Hence, the monitoring of EEG with human intervention is a complicated task. Hence, reliable identification of seizure prevents the supervising process and investigation of seizures. Moreover, the contribution of medical workers during seizure activities is more important for analyzing patient having seizures. The frequent communication by monitoring patient helps to discover the pertinent possessions of seizure that involves sternness and introduce embellished region of the brain [11]. Hence, automatic discovery of seizure helps to aware clinicians in initiating seizures in which the instant therapeutic intervention can arise [12,2].

Various automatic detection methods are discovered in the literary works that make efforts of neurologists simpler and quicker [16]. These techniques involve preprocessing acquisition of imperative features, and classification. The actions of preprocessing involve filtering and elimination of noise from the input signals of EEG and are crucial to enhance the effectiveness of the technique. The extraction of features is the next step to distinguish the seizures and non-seizure based on the efficiency of the classifier. The frequency and time domain are extensively used features for performing automatic seizure detection. Several features are adapted for seizure detection which involves entropy, mean, fractal intercept, lacunarity, power spectral density, energy Hurst component, zero crossings, and Renyi entropy [17]. Also, various classifiers, such as K-nearest neighbor (K-NN), linear discriminate analysis (LDA), machine learning, artificial intelligence, support vector machine (SVM), and fuzzy c-means which are utilized to classify the problems of epileptic seizures [18]. There exist several issues that are solved for detecting seizures considering time-frequency techniques with an off-line database [15]. The analysis of seizure in real-time for instantaneous therapeutic explication is tremendously obligatory and the prologue of automatic biomarker enhances the remedial verdict. The reliability of seizure detection depends on pertinent feature selection using recordings of EEG [3].

The purpose is to present a technique for discovering epileptic seizures with EEG. The major part is to design an optimization method for detecting seizures. Here, the signal is obtained from a database and applied to the pre-processing step for removing the artifacts. Afterthat, the pre-processed EEG is divided into different channels for extracting features with each channel. The features, such as tonal power ration, relative energy, spectral features, fluctuation index, LPC, Taylor-based delta AMS, and Holoentropy are obtained with each channel. The obtained features from each channel are constituted in the feature vector and are the combination of features. The size of the feature vector is reduced using PPCA. The feature vector with reduced dimensionality is categorized with Deep SAE. The Deep SAE undergoes training using proposed Adaptive rag-ROA and is designed by combining adaptive ideas in rag-ROA. Hence, the Adaptive rag-ROA-based deep SAE is employed to detect seizures from EEG.

The major contributions are listed below:

• Proposed Adaptive rag-ROA-based Deep SAE for discovering elliptical seizure: The proposed Adaptive rag-ROA based deep SAE is utilized wherein the proposed Adaptive rag-ROA is adapted to train deep SAE. Here, the integration of the Adaptive idea in rag-ROA is done for seizure detection.

Other sections are organized as given below: Section 2 portrays elaboration of classical elliptical seizure discovery strategies used in literary works and issues confronted that are employed as motivation for devising the proposed method. The proposed technique for detecting elliptical seizure considering Deep SAE is elaborated in Section 3. The analysis is elaborated in Section 4 and finally, Section 5 offers a conclusion.