3-Layer LSTM Model for Detection of Epileptic Seizures

An electroencephalogram (EEG) is one of the ancillary methods to record the signals generated by the electrical activity of the brain. Conventionally, neurologists scrutinize these EEG signals to identify neurological abnormalities such as epilepsy. Such a way of observation is too time-consuming and requires proficiency. Therefore, a computer-aided diagnosis (CAD) system is needed to discriminate the class of these EEG signals automatically. This paper employs long short-term memory (LSTM) for the analysis of EEG signals. Herein, the LSTM model having only three layers is presented. This model achieved 98.5% accuracy to differentiate between non-seizures and seizures only in 30 epochs. Less number of layers and epochs are the main attraction of this work, which makes this model useful for real-time detection purpose.