Biomedical Engineering Reference
In-Depth Information
Their magnitude and visibility in the EEG depends on various factors including the
EEG reference and montage used [16]. Normally, removal of the EKG influence is
not required because it is rarely time-locked to the stimulation and thus will average
out during EEG processing. However, if a profound affliction makes EKG artifact
removal necessary, this can be done either with a template subtraction based
approach or with ICA. A different artifact related to this is the pulse artifact caused
by the placement on an electrode directly above a blood vessel, resulting in pulsatile
artifacts at the heartbeat frequency, a problem that is particularly present in simulta-
neous EEG-fMRI recordings [17]. Another source of endogenous EEG artifacts is
the respiratory system, which can cause slow variations in scalp impedance resulting
in equally slow shifts. Finally, sweating of the scalp can have a profound effect on
the EEG since the sodium chloride and other sweat components such as lactic acid
react with the electrode metals to produce battery potentials that present in the EEG
as slow oscillations (0.1 to 0.5 Hz) of fairly large magnitude.
Figure 2.6 shows examples of three endogenous and one exogenous artifact.
Figure 2.6(a) shows a vertical eye blink and how the channels are affected by this to
differing degrees. Blink artifacts are quite large in amplitude with typical blink peak
values being on the order of several hundred microvolts. Figure 2.6(b) shows a typi-
cal horizontal eye movement, with the lateralized activity on the electrodes at the
outer canthi of the eyes being clearly visible in the lower middle part of the panel.
Both the vertical as well as the horizontal eye movements can easily be detected
based on their unique topographies and can subsequently be removed, for example,
with a regression-based or, better, an ICA-based method (see Section 2.4). Muscle
activity related artifacts, as shown in Figure 2.6(c), typically contaminate the EEG
at higher frequencies. EMG artifact reduction is often based on the application of a
lowpass filter. Yet to some extent the EEG frequency range of interest may overlap
with the broadband contamination that muscle activity contributes, making this
artifact notoriously difficult to remove. In addition to the endogenous artifact
examples shown in Figure 2.6(a-c), a range of exogenous artifacts can occur as
well. The spike-like example shown in Figure 2.6(d) can be classified as an exoge-
nous artifact event, because it is spatially restricted to a single channel, making a
brain source highly unlikely due to the lack of volume conductance-related spatial
smearing. Indeed, the better spatial sampling of state-of-the-art EEG recordings
improves the identification and characterization of exogenous as well as endoge-
nous artifacts.
2.3.3 Artifact Removal Techniques
Artifact rejection is required when the artifacts present in the EEG data cannot be
removed algorithmically. For this, a great range of different measures and
approaches exists, including these:
Simple amplitude threshold:
This value defines positive and negative ampli-
tude levels above/below which data is automatically recognized as artifacts.
Min-max thresholds:
This measure sets a maximally allowed amplitude differ-
ence within a specific length of time. This achieves something very similar to
Search WWH ::
Custom Search