Two-layer LSTM network-based prediction of epileptic seizures using EEG spectral features

Abstract Epilepsy is a chronic nervous disorder, which disturbs the normal daily routine of an epileptic patient due to sudden seizure onset. In this era smart healthcare, automated prediction techniques could assist patients, their family, and medical personnel control manage these seizures. This paper proposes spectral feature-based two-layer LSTM network model for automatic seizures using long-term multichannel EEG signals. makes use power mean spectrum amplitude features delta, theta, alpha, beta, gamma bands 23-channel task. Initially, proposed single-layer models have been evaluated segments having durations in range 5–50 s 24 subjects, out 30 duration are found be useful accurate model. Afterwards, validate performance classifier, fed random forest, decision tree, k-nearest neighbour, support vector machine, naive Bayes classifiers, empowered with grid search-based parameter estimation. Finally, iterative simulation results comparison recently published existing firmly reveal that effective technique accurately predicting real time average classification accuracy 98.14%, sensitivity 98.51%, specificity 97.78%, thereby enabling patients better quality life.