Application to biomedical signals - Practical Biomedical Signal Analysis Using MATLAB - page 144

deflection and, consequently, less distorted by time-domain averaging. For longer
latency of brain response, latency jitter is greater, so short-lasting components with
long latency can hardly by observed [Mouraux and Iannetti, 2008].
We can assume that in equation 4.11 the variability of the signal is due only to
the variable latency and for each trial i the shape of the response s ( t ) is invariant. It

implies that there exists a time translation Δ t i such that:

( )= ( + ) .

s i t s t Δ t i (4.14)

The first approach to correction for the latency variation was proposed by Woody
[Woody, 1967]. In his method theΔ t i was estimated by means of the cross-correlation
function of the response i with a template (i.e., the initial estimate for the average EP).
The required translation was evaluated as the lag of maximum of cross-correlation.
The individual trials are then readjusted byΔ t i and averaged, giving the new estimate
of average EP. This estimate was used as a template in the next step of the process.
This iterative procedure leads to a better estimate of the average EP. Woody showed
also that the magnitude of the cross-correlation function can be used to sort the single
trial EP to different sub-classes. The refinement of Woody's method relied on taking
into account the statistical properties of the ongoing EEG activity [McGillem and
Aunon, 1977].
A method of latency correction, involving additionally a classification step, was
proposed by [Pfurtscheller and Cooper, 1975]. According to this technique sin-
gle trial EPs are cross-correlated with different templates. The maxima of cross-
correlation functions are used to estimate the amplitude and latency of particular
components. Next, these trials that fall within the certain amplitude or latency limits
are averaged together. In this way, different classes of selected averages are obtained
that may correspond to different subsystems or physiological conditions. Averaging
of all trials as one class may lead to a considerable information loss.

4.1.7.1.4 Habituation The observed trial-to-trial variability in amplitude and la-
tency of the components of an evoked potential may arise from two major sources.
One is systematic and depends on the changes in reaction to the repetitive stimula-
tion, and the other is stochastic. Systematic changes in response to repeated stimula-
tion are a general property of a nervous system. New stimuli first elicit an arousing
response, named the orientation response [Sokolov, 1960]; it consists of changes in a
large number of autonomic variables, such as the skin conductance reaction, phasic
heart rate changes, and it also involves a reaction of the cerebral cortex. Habituation
is, in a wide sense, defined as decrease of response as a function of stimulus rep-
etition and it is a constant finding in almost any behavioral response. The simplest
mathematical model of habituation is an exponential decay in case of regularly re-
peated stimulus [Sokolov, 1960]. The rate of decay depends on physical properties of
the stimulus, its relevance, the interstimulus interval, and the properties of the neural
systems involved in the reaction. Habituation is not the only possible reaction to a
sequence of stimuli. The increases in the responses during the first stimuli related to
a sensitization process have been described in the literature [Thompson and Spencer,