EEG signal pre-processing for segmentation into significant regions, major artefacts removal, and uncertainty reduction in epileptic seizure characterization

Abstract

EEG analysis is widely used for clinical investigations of several neurological disorders. As EEG signal is normally a low amplitude (microvolts) signal and recording is multichannel, with epochs lasting from several minutes to hours depending on the test focus, difficulties can arise due to artefacts contaminating the data and also due to the presence of various events in the signal that could occur frequently. This is why, in order to improve the accuracy of clinical conclusions based on EEG analysis, it is important to provide the scientific research and then the clinical routine with EEG prepocessing methods aiming to remove major artefacts, perform suitable segmentation of EEG for further analysis and for accurate characterisation of events detected in segmented epochs such as the epileptic seizure event reducing the uncertainty in clinical analysis. Here we describe several methods for that goal.