Proficient RO-RNN Learning Model for Seizure Prediction Systems

Abstract

According to WHO, world widely around 65 million people are affected by epilepsy. Prediction of such a lifethreatening neurological disease is of high importance. Predictability of seizures uplifts the patient’s life and
wellbeing. This paper presents the application of machine learning in the prediction of epileptic seizures. In this
work, we used Regularized and Optimized Recurrent Neural Network (RO-RNN). The aim of this work is to
investigate the application of bidirectional long short-term memory (LSTM) networks for epileptic seizure
prediction. A Weight Dropped (WD) method is used for regularizing the LSTM model and an Averaged
Stochastic Gradient Descent (ASGD) is used for optimizing the LSTM model. Regularization and optimization
are deployed with the deep neural network architecture to accelerate the convergence rate and to reduce the
complexity of the proposed non-linear model. The proposed model is evaluated using two diverse public
databases such as traditional CHB-MIT and recent NINC respectively. Also, a private real time SRM database is
used for the assessment of the proposed computer-aided seizure prediction approach. Empirical results on 200
recordings outperforms the state-of-art approaches with an accuracy score of 0.91, sensitivity score of 0.89 and
false prediction rate of 0.12 FP/h. Experimental results prove that the proposed seizure prediction approach is a
promising one for accurate real-time prediction of epilepsy using scalp EEG data.