Magnetoencephalography (MEG) - Dynamic Brain Imaging - page 170

V1 modulation well within 100 ms by activity in the amygdala
some 40-50 ms earlier (4, 31, 32) . This early amygdala and V1
activations and their interaction were identified in the responses
elicited by emotional faces, providing support for the existence
of a low route to the amygdala as postulated by Le Doux
(33) . In addition to the obvious theoretical importance of this
observation for normal brain function, it was also found that
early and late amygdala and V1 activations, and the interaction of
these and other areas (fusiform gyrus and inferior frontal cortex)
were different in normal subjects and schizophrenics (31, 32, 34) .
These MFT findings were recently reproduced with remarkable
consistency in the timing of amygdala and V1 activity as an
event-related synchronization in response to fearful faces using
beamformer techniques (35) .
We conclude this section with examples demonstrating how
the nature of the output changes as we move from MEG signals
to tomographic estimates of activity, and measures that describe
the properties of groups of trials. We will show examples that
progress systematically in complexity using either single trial data
or the average of a small number of trials to mine the MEG data.
The first example, Fig. 8.1 , shows how strong features can be
extracted directly from a signal signature, almost a direct reading
of the MEG single trial signals. The data were collected from
a median nerve stimulation that was strong enough to elicit a
finger twitch. The first 11 responses are shown by the triggers
collected simultaneously with the MEG signal ( Fig. 8.1a ).
This unusually strong stimulus and the superficial location of
the primary somatosensory cortex (S1) make the first cortical
responses elicited by each stimulation almost visible in the raw
signal. A virtual sensor can be constructed in this case (see below)
which enhances the pattern of signal produced by S1 while
reducing other signals. Despite this enhancement, the actual
evoked response is barely discernible in the presence of other
similar activations ( Fig. 8.1b ). The evoked response becomes
easier to see in the zoomed version covering the third and fourth
median nerve stimulations ( Fig. 8.1c ). In this figurine, two
more properties of the MEG signal are evident. First, the evoked
response is similar from trial-to-trial in its slow envelope but
the fast transients riding on it are highly variable from trial to
trial. Second, the virtual sensor (VS) output for the signal from
the brain is always much higher than the system noise, a direct
demonstration of the high dynamic range of MEG. The VS
for the system noise is the barely distinguishable trace hugging
the zero level horizontal axis; it is obtained by processing the
MEG signal recorded when no subject is in the shielded room
in exactly the same way as the signal from the subject. The
VS is constructed by taking the difference of the means of the
7 strongest positive and 7 strongest negative MEG channels