Techniques of EEG Recording and Preprocessing - Quantitative EEG Analysis Methods and Clinical Applications - page 25

which is a noise estimate that is less dependent on the validity of the additive ERP
model assumptions.
2.1.2 Event-Related Oscillations
Continuous EEG recordings consist largely of oscillations at different frequencies
that fluctuate over time and provide valuable information about a subject's brain
state. Brain oscillations such as EEG alpha activity (8 to 13 Hz) clearly respond to
sensory stimulation (e.g., alpha suppression). To what extent these oscillations con-
tribute to event-related EEG signals such as ERPs is a matter of ongoing research [3,
4]. Notwithstanding this discussion, it has become evident that the ERP does not
necessarily capture all event-related information present in the EEG. For instance,
oscillations induced by, but not perfectly phase-locked to, an event of interest zero
out in the process of ERP calculation. It is therefore helpful to distinguish between
evoked, phase-locked oscillations and induced, nonphase-locked signals. In this
context, the term total power refers to the sum of evoked and induced oscillations.
Total power is calculated by summing the values of the frequency transform of the
single trials, whereas evoked power is obtained by a frequency transform of the
time-domain average, namely, the ERP.
A change in oscillatory power can be due to a change in the size of the neuronal
population generating the oscillation, or it can reflect a change in the degree of syn-
chronization of a given neuronal population. With the latter mechanism in mind,
Pfurtscheller coined the expression event-related synchronization (ERS) for relative
power increases and event-related desynchronization (ERD) for relative power
decreases [5]. ERS/ERD calculations can be displayed over time (e.g., by using a
short-time fast Fourier transform or wavelet decomposition) and expressed as a per-
centage signal change relative to the pre-event reference period. For instance, the
amount of EEG alpha activity prior to the presentation of a visual target predicts, to
a substantial extent, whether the target will be consciously perceived [6, 7]. More
generally, the ERS/ERD type of analysis and its extension to the broad frequency
range [8] reveals important information about brain function without assuming
independence between ongoing activity and brain-electrical responses.
2.1.3 Event-Related Brain Dynamics
The distinction between evoked (ERP) and induced (event-related oscillations)
brain activity suggests that, beyond the consideration of power changes as in the
original ERS/ERD analysis, the consistency of phase across epochs provides rele-
vant information. Indeed, ERPs could also be the result of changes in the phase con-
sistency of ongoing oscillations in the absence of a power increase [3], a
phenomenon that has been named partial phase resetting (PPR). To investigate
these mechanisms, it is necessary to move from the analysis of averaged brain
response to the analysis of single epochs or trials. That is, the basis for the consider-
ation of phase concentration is the frequency, or time-frequency, analysis of every
single recorded trial. This is expressed in the event-related brain dynamics model
[4], which represents a three-dimensional signal space with the axes power change,
frequency, and phase consistency (Figure 2.2).