Visualizing local linear embedding and fast ICA with linear neural networks for epilepsy classification

Abstract

Characterized by unexpected and sudden electrical disturbances in the brain, this neurological disorder termed epilepsy is a serious problem in the medical world. When the cortical neuronal networks synchronize excessively, then such electrical disturbances takes place which leads to a seizure. The details about all the functions of the brain can be obtained easily with the help of Electroencephalography (EEG) signals. Generally, the recordings of the EEG signals are too long and so processing it are quite difficult. In this paper, Local Linear Embedding (LLE) and Fast Independent Component Analysis (ICA) are utilized as dimensionality reduction techniques. The dimensionally reduced values are then classified with the help of Linear Neural Networks for the classification of epilepsy from EEG signals. The results show that when LLE and Fast ICA are used as dimensionality reduction techniques and classified with Linear Neural Networks, an average classification accuracy of 97.51% is obtained with LLE and 97.87% is obtained with Fast ICA.