On Digitally Identifying the Very low-lying Amplitude Deviation and Rhythmic Irregularities in EEG Signal for Epileptic Seizure Detection and Analysis

Abstract

Purpose: Recent studies signify that epilepsy has become significant burden to the society due to its unpredictable nature and high cost of treatment. It has been observed that, an integrated neural investigation system is required to aid epileptologist for early diagnosis as well as long-term monitoring of patients. Electroencephalogram (EEG) plays a significant role to study brain’s electrical activity such as epilepsy. An epileptic EEG signal have four necessary distinctive characteristics- sudden changes in signal, amplitude deviations, electro-cerebral negativity, and physiological fields formation by affected EEG channels.
Materials and Methods: In this paper, we proposed an automated system for epileptic seizure prediction using above mentioned characteristics. In the proposed system, the very low-lying amplitude deviation and rhythmic irregularities were identified in EEG signal. The proposed work is based on extraction of features of EEG signal and structure of brain. The proposed system provide dual execution mode ( detection and analysis ) and is able to predict the epileptic seizure waveforms and specific brain regions that will be affected by upcoming seizures.
Results: The proposed system accurately identifies various seizure waveforms in EEG signals of primary dataset and achieves an excellent performance (92.66% accuracy, 94.86% F1 score) in prediction of epileptic seizures and in case of secondary dataset it show 100% accuracy.
Conclusion: Advantage of the proposed system lies on generation of a detailed report on upcoming seizure location and nature prior to the seizure which mitigate patient life risks. This advancement could lead to substantial improvements in the field of neuroscience.