Biomedical Engineering Reference
In-Depth Information
subject performed a visual-spatial selective attention task during which he covertly
attended one of five squares continuously displayed on a black background 0.8 cm
above a centrally located fixation point. Four squares were outlined in blue, and
one, marking the attended location, was outlined in green. The location of this
green square was counterbalanced across 72-second trial blocks. The subject was
asked to press a right-hand-held thumb button as soon as possible following target
stimulus presentations in the attended location (the green square), and to ignore the
similar (nontarget) stimuli presented in the other four boxes. Stimuli were white
disks, presented in one of the five boxes at random. EEG data was collected from 29
scalp electrodes mounted in a standard electrode cap (Electrocap, Inc.) at locations
based on a modified International 10-20 system, and from two periocular elec-
trodes placed below the right eye and at the left outer canthus. Data was sampled at
512 Hz (downsampled to 256 Hz) with an analog pass band of 0.01 to 50 Hz.
Although the subject was instructed to fixate the central cross during each block, he
tended to blink or move his eyes slightly toward target stimuli presented at
peripheral locations.
After ICA training on 555 concatenated 1-second data trial epochs, independ-
ent components that accounted for blinks and eye movements were identified by the
procedures detailed in [43] based on the characteristics of time course of component
activations, the component scalp topographies, and the locations and orientations
of equivalent dipoles obtained using functions available in the freely available
EEGLAB environment [44]. Here, ICA successfully isolated blink artifacts to a sin-
gle independent component [Figure 2.9(a)] whose contributions were removed from
the EEG record by subtracting its component projection from the data.
Though the subject was instructed to fixate the central cross during each block,
the technician watching the video monitor noticed that the subject's eyes also
tended to move slightly toward target stimuli presented at peripheral locations. A
second independent component accounted for EEG artifacts produced by these
small horizontal eye movements [Figure 2.9(b)]. Its scalp pattern is consistent with
that expected for lateral eye movements. Note the overlap in scalp topography
between the two independent components accounting for blinks [Figure 2.9(a)] and
for lateral eye movements [Figure 2.9(b)]. Again, unlike PCA component maps, ICA
component maps need not be orthogonal and may even be nearly spatially
coincident.
A standard approach to ERP artifact rejection is to discard eye-contaminated
trials containing maximal potentials exceeding some selected value (e.g.,
V) at
periocular sites. For this dataset, this procedure rejected 78 of 555 trials, or 14% of
the subject's data. Figure 2.9(c) shows ERP averages of relatively uncontaminated
target trials (solid traces) and of the contaminated target trials (faint traces) that
would have been rejected by this method. These averages differ most at frontal elec-
trodes. Figure 2.9(d) shows averages of the same uncontaminated (solid traces) and
contaminated (solid traces) trials after the independent components accounting for
the artifacts were identified and removed, and the summed activities of the remain-
ing components projected back to the scalp electrodes. The two ICA-corrected aver-
ages were almost completely coincident, showing that ICA-based artifact removal
did not change the neural signals that were not contaminated. Note that
the ICA-corrected averages of these two trial groups are remarkably similar to the
=
60
μ
Search WWH ::
Custom Search