EARLY SEIZURE DETECTION TECHNIQUES: A Review

Abstract

Objectives: To study the different seizure detection techniques and find an efficient technique which results in optimum output parameters in terms of sensitivity and specificity. Methods/Statistical Analysis: Different seizure detection techniques were studied to find the efficient technique for early seizure detection. Past researches extracted different features from raw Electroencephalogram (EEG) signals. Features like time domain and frequency domain were extracted. Recently, wavelet parameters have been introduced which is the combination of time and frequency domain. For the classification, different classifiers like RNN, Artificial Neural Network, Modified Neural Network and Support Vector Machine were used. Findings: This review paper studies different types of seizure detection techniques. Development of early seizure detection techniques gains a major attention to provide an early alert to the epileptic patient. Different time domain, frequency domain and time-frequency domain features were extracted from the raw EEG signal. For the classification, different classifiers like RNN, Artificial Neural Network (ANN), Modified Neural Network and Support Vector Machine (SVM) were used. It was observed that the technique employing DT-CWT as a feature set and Support Vector Machine as a classifier resulted in maximum classification accuracy of 100% and 0 false alarm rate. So, from the findings, it is concluded that the combination of wavelet decomposition and a number of features extraction will improve the accuracy for early seizure detection in future. Application/Improvements: In future, the early seizure detection techniques can be used to provide an alert to the epileptic patient so that an effective diagnosis can be done before the occurrence of seizure onset.