Epileptic Seizure Detection Using LSTM: A Deep Learning Technique

Abstract

Epilepsy is one of the most devastating diseases in the history of mankind. It is a neurological disorder in which irregular transmission of brainwaves results in seizures, physical, and emotional imbalance. This paper presents the study of the effective use of deep learning algorithms in epileptic seizure detection. This innovative technology will help us to make a diagnosis of the disease faster and with greater accuracy. We are provided with a feature-extracted dataset consisting of 179 attributes. The original dataset consists of five folders with 100 files, each one representing a single person. Initially, binary classification is done by considering the class with epileptic activity against all other classes. For classification, we used the LSTM architecture model and landed above accuracy of over 97%. Then by applying a different multiclass dataset having five classes, we checked the degree of generalization of our models and the accuracy of end results was high. We also generated copies of the model and applied ten-fold cross-validation methods on our models for better performance. All the results are shown with the help of confusion matrices and accuracy-loss plots.