Epileptic Seizure Classification using LSTM

Abstract

Epilepsy is a neurological disease which is non transmissible and affects across all ages. It is signalized by observing repetitions of seizures. Main causes of epilepsy are head injuries, low oxygen during birth, brain tumors, genetic conditions and infections like meningitis or encephalitis. A seizure came out whenever explosion of electrical impulses in brain elude their normal restricts. Traditional diagnosis techniques may take time to diagnose and it may differ in perfection as compared to automated system due to lack of expertization in each case. So this is a serious issue that disorder must be diagnose prior to the manifestation of behavioral symptoms. The main motive of this study is bestow expert model for recognizing disorders on basis of Electroencephalogram (EEG) data using Deep learning methodology. In this paper to detect epileptic seizures, author presented long short-term memory (LSTM) model on Bonn’s EEG dataset. Metrics calculated to check the efficacy of the technique are accuracy, specificity and sensitivity. Proposed model has achieved 98.5% accuracy, 99% sensitivity and 98% specificity only in 30 epochs.