Biomedical Engineering Reference
In-Depth Information
latency axis. We compute SP, NP and SNR as described elsewhere
(36)
and give the relevant formulae in Appendix 2.
Figure 8.2b
shows the SP, NP and SNR for the six trials in the same order as
part (A) of the same figure. It is evident that using only six trials
in each set is enough for a faint response pattern to be seen in the
average (heavy black line in
Fig. 8.2a
) and clear strong peaks in
the SNR computed using the 20 ms sliding windows across the
six single trials (
Fig. 8.2b
). The high SNR values are computed
from SP and NP measures that show a much smoother behavior
than the raw activation curves. It is evident in
Fig. 8.2b
that the
peaks in SNR values are sometimes due to increase in SP; at oth-
ers, due to decrease in NP and, at yet others, due to a combination
of the two. The message from
Fig. 8.2
is that the use of measures
that exploit the information in single trial tomographic analysis
allows reliable extraction of information from a small number
(6)
of trials. Information of similar quality is available in the average
signal only when a large number (typically hundreds) of trials are
averaged. The average signal of large number of trials shows peaks
that cluster around the SNR peaks of regional activations, often
from more than one area. These results have significant implica-
tions for experimental designs. In general, it is highly desirable to
have only a few trials of a given type within a session, for exam-
ple to avoid habituation, or to include many different conditions
in each run. Since this is incompatible with averaging hundreds
of trials, the design of most experiments becomes a choice of
lesser evils. The capability of single trial tomographic analysis of
each timeslice therefore allows for more powerful experimental
designs, as the next example demonstrates.
The last example summarizes the results of a recent study of
illusory contour perception with stimuli presented at the cen-
ter of the visual field and in each of the quadrants
(37)
.Using
tomographic analysis of average data from relatively few trials in
each run, we were able to study in the same experiment and for
each part of the visual field, the effect of attention as reflected by
different task demands. The MFT solutions were used to com-
pute activation curves mainly within the primary visual cortex
(V1/V2), the lateral occipital cortex (LOC) and FG for each
subject, visual field presentation, condition and stimulus type.
The location of the V1/V2, LOC and FG ROIs identified for
each subject were transformed into the space of Talairach and
Tournoux
(38)
and finally back-projected onto the MRI of one
subject for display purposes. The statistical analysis of the acti-
vation curves from seven subjects demonstrated that a different
mechanism operates for the processing of illusory figure process-
ing when the stimuli are presented in the center and periphery of
the visual field
(37)
. The results are displayed in
Figure 8.3
for
stimuli presented in the center (
Fig. 8.3a
) and for stimuli pre-
sented in the four quadrants (
Fig. 8.3b
) of the visual field. In
