A Convolutional Neural Network for Artifacts Detection in EEG Data

Abstract

AbstractElectroencephalography (EEG) is an effective tool for neurological disorders diagnosis such as seizures, chronic fatigue, sleep disorders, and behavioral abnormalities. Various artifacts types may impact EEG signals regardless the used, resulting in an erroneous diagnosis. Various data analysis tools have therefore been developed in the biomedical engineering literature to detect and/or remove these artifacts. In this sense, deep learning (DL) is one of the most promising methods. In this paper, we develop a novel method based on artifacts detection using a convolutional neural network (CNN) architecture. The available EEG data was collected using 32 channels from the Nihon Kohden Neurofax EEG-1200. The data are preprocessed and analyzed using our CNN to perform binary artifact detection. The suggested method highlights the best classification results with a maximal accuracy up to 99.20%.