Early Study in Automatic Identification of Epilepsy in Neonatal Using EEGLAB and One Dimensional Convolutional Neural Network Through the EEG Signal

Abstract

In detecting epileptic activity, medical experts examine the visual result of Electroencephalography signals. The visual analysis will take a lot of time and effort, due to a large amount of data. Furthermore, there are some errors in concluding the analysis result. One of the ways to analyze this quickly is to use Machine Learning (ML) methods. This study aims to evaluate the performance of 1D-CNN in identifying the given data. First, the signal will go through pre-processing using EEGLAB Toolbox which is then classified to identify epilepsy and non-epilepsy with the 1D-CNN algorithm. The results showed that the proposed method obtained high accuracy values, respectively 99,078% for the training data and 82,069% for the validation results. From the evaluation by a confusion matrix, an average accuracy of 99,31% was obtained. Based on this evaluation, the proposed model can be used as an efficient method in the process of automatic classification, detection, or identification of epileptic activity.