A Reliable Statistical Method to Detect Eyeblink-Artefacts from Electroencephalogram Data Only

Abstract

For data preprocessing and artefact removal in an ERP experiment we were confronted with the question how blink artefacts can be detected reliably, even in the absence of usable electrooculogram (EOG) data. We propose an objective and quantitative method for the automatic detection of eyeblink artefacts from raw data using extreme value statistics, with a p-value acting as a threshold parameter. For testing the method, we used 29 channel electroencephalogram recordings of 55 healthy subjects. A total 7,700 s of EEG were analysed. The proposed method was found to detect blink artefacts reliably, showing that extreme value statistics can be employed to detect blink artefacts, even in the absence of EOG recordings.