Effect on EEG responses of removing ocular artefacts by proportional EOG subtraction

Work undertaken to investigate the effect on EEG responses of ocular artefact removal by proportional electro-oculogram subtraction using the least-squares method is reported. The ocular artefact model and the least-squares approach to estimation of the model parameters and hence of the EEG waveform, including the response, are described. Results are presented to show that the response shapes become modified if ocular artefact removal is implemented recursively. This happens because the response is incorrectly treated as part of the random background EEG. The solution is to incorporate a model of the response within the ocular artefact removal algorithm. The results of tests on simulated data confirm these conclusions. The results of the incorporation of response models in removing ocular artefacts from CNV recordings are presented. Even with modelling recursive ocular artefact removal is found to modify the shape of the CNV, and so reliable results can only be obtained if the removal is implemented nonrecursively. Evidence is given that it is advisable to remove DC levels from the recorded data.