Automatic removal of eye movement artifacts from the EEG using ICA and the dipole model
Abstract
In this study,we proposed and evaluated the use of Independent Component Analysis(ICA) combining the EEG dipole model to automatically remove eye movement artifacts from the EEG without needing EOG as a reference.
We separated the EEG data into independent components using the ICA method,and determined the source localization of these independent components with a single dipole model.
The EEG signal was reconstructed by antomatically excluding those components localized within a preset eye model.
The experimental results indicate that the dipole model is very efficient at automatically
substracting the eye movement artifacts,while retaining the EEG slow waves and making their
interpretation easier.
Methods Comparison
other methods’s disadvantages:(filter、recording of horizontal and vertical EOG)
- used a simple filtering concept,simply ignoring very low frequencies(below 1.5 or 2 Hz)
- would not perform well in the context of a large amount of frontal slow waves (EOG recordingwould be contaminated by this slow wave activity)
- The ICA algorithm is computationally efficient.
- better noise suppression ability.
- ICA can simultaneously separate the EEG and artifacts into independent components without relying on the availability of reference artifacts.
- avoids the problem of mutual contamination between EEG and EOG channels that could not be solved with filters,regression and PCA.
- The corrected EEG can easily be derived by a combination of the components without artifacts.
A crucial step for ocular artifact correction using ICA algorithms:
to correctly identify the artifact components among the decomposed indepent components.
Manually identifying:
- subjective
- inconvenient
- time consuming