EEG-based epileptic seizure detection using GPLV model and multi support vector machine

Abstract

Epilepsy is chronic neurological disorder that is clinically detected by continuous monitoring of Electroencephalogram (EEG) signals by experienced clinicians. Epilepsy is detected by clinicians based on the visual observation of EEG records that normally consumes more time and sensitive to noise. To address these issues, a new system is proposed for automatic epileptic seizure recognition. Gaussian Process Latent Variable Model (GPLVM) is used to lessen the number of features by achieving a set of principal features that significantly rejects the “curse of dimensionality” concern. Then, a supervised classifier (Multi Support Vector Machine (MSVM)) is used to classify the epileptic seizure classes such as normal, ictal, and interictal. Experimental result exemplifies that the proposed work effectively classifies the epileptic seizure classes in light of sensitivity, specificity, False Positive Rate (FPR), False Negative Rate (FNR), and accuracy. The proposed work improves the classification accuracy upto 2.5-12% related to the existing works. The proposed work delivers a new avenue for assisting clinicians in diagnosing epileptic seizure.