A comparison of ocular artifact removal methods for block design based electroencephalography experiments

Reinmar Kobler, Andreea Ioana Sburlea, Gernot Müller-Putz

Research output: Chapter in Book/Report/Conference proceedingConference paperpeer-review


Eye movements and their contribution to electroencephalographic (EEG) recordings as ocular artifacts (OAs) are well studied. Yet their existence is typically regarded as impeding analysis. A widely accepted bypass is artifact avoidance. OA processing is often reduced to rejecting contaminated data. To overcome loss of data and restriction of behavior, research groups have proposed various correction methods. State of the art approaches are data driven and typically require OAs to be uncorrelated with brain activity. This does not necessarily hold for visuomotor tasks. To prevent correlated signals, we examined a two block approach. In a first block, subjects performed saccades and blinks, according to a visually guided paradigm. We then fitted 5 artifact removal algorithms to this data. To test their stationarity regarding artifact attenuation and preservation of brain activity, we recorded a second block one hour later. We found that saccades and blinks could still be attenuated to chance level, while brain activity during rest trials could be retained.
Original languageEnglish
Title of host publicationProceedings of the 7th Graz Brain-Computer Interface Conference 2017
Subtitle of host publicationFrom Vision to Reality
EditorsGernot Müller-Putz, David Steyrl, Selina Wrissnegger, Reinhold Scherer
Place of PublicationGraz
PublisherVerlag der Technischen Universität Graz
Number of pages6
ISBN (Print)978-3-85125-533-1
Publication statusPublished - 18 Sept 2017
Event7th Graz Brain-Computer Interface Conference 2017 -
Duration: 18 Sept 201722 Sept 2017


Conference7th Graz Brain-Computer Interface Conference 2017


  • Eletroencephalography
  • Ocular artifacts
  • Independent component analysis
  • regression

Fields of Expertise

  • Human- & Biotechnology

Cite this