Biomedical Engineering Reference
In-Depth Information
EEG signal
Digitizing
Artifact
filtering
Suppression
detection
Fast Fourier
transform
Bispectrum
BSR and
QUAZI
SynchSlow
Beta ratio
BIS = Weighted sum of subparameters
Figure 9.9
Overview of BIS algorithm.
ond duration. A series of algorithms then detects and attempts to remove or ignore
artifacts.
The first phase of artifact handling uses a cross-correlation of the EEG epoch
with a template pattern of an ECG waveform. If ECG or pacer spikes are detected,
they are removed from the epoch and the missing data estimated by interpolation.
Epochs repaired in this phase are still considered viable for further processing.
Next eye-blink events are detected, again relying on their stereotypical shape to
match a template with cross-correlation. Epochs with blink artifacts are considered
to have unrepairable noise and are not processed further. Surviving epochs are
checked for a wandering baseline (low-frequency electrode noise) and if this state is
detected, additional filtering to reject very low frequencies is applied. In addition,
the variance (i.e., the second central moment) of the EEG waveform for each epoch
is calculated. If the variance of an epoch of raw EEG changes markedly from an
average of recent prior epochs, the new epoch is marked as “noisy” and not pro-
cessed further; however, the new variance is incorporated into an updated average.
If the variance of new incoming epochs continues to be different from the previous
baseline, the system will slowly adapt as the prior average changes to the new
variance.
Presuming the incoming EEG epoch is artifact free, or is deemed repaired, the
time-domain version of the epoch is used to calculate the degree of burst suppres-
sion with two separate algorithms: BSR and QUAZI [6]. The BSR algorithm used by
the BIS calculation is quite similar to that described in the preceding section. The
Search WWH ::
Custom Search