Biomedical Engineering Reference
In-Depth Information
signals evoked by stimuli
(22)
and epileptic activity
(23)
,al
demonstrated the ability of MFT to identify distinct foci of
activity as their strength changed while ECD models described
them as wandering dipoles through the brain. In the last ten
years, accurate MFT reconstructions have been demonstrated
with many applications and with the emphasis recently placed
on reconstructions of real-time data, i.e. for single timeslices of
single trials
(4, 14)
.
4. Output of MEG
Neuroscience has, at its disposal, a plethora of exquisite tech-
niques to study neural activity. In the vast majority of cases,
the output of techniques is sufficiently constrained to limit the
choices of the researcher to qualitatively one distinct category of
output. The researcher has to perfect the technique to obtain the
data of the highest possible quality, but she/he cannot change
the qualitative nature of the neural events she/he is studying.
For example, in a microstimulation experiment, one may worry
about how large an area is excited but, one is certain that what
is examined is the disruptive effect of injecting a current that
perturbs the local neural interactions. Similarly, when one uses
fMRI or PET, one can safely assume that one measures correlates
of neural function mediated by blood supply and, hence, over
delays of seconds. With MEG, the case is somewhat different
because the method allows one to focus either at fine detail in
space and time, and/or within a small area, or across the brain
at timescales ranging from a fraction of a millisecond to hours.
Quantitative changes in the choice of what spatiotemporal scales
to consider imply sensitivity to qualitatively different neuronal
events and organization. Using MFT, for example, to image in
real time, one can follow changes in the brain at a fraction of
a millisecond
(14)
, i.e. at a temporal resolution that is about
one order of magnitude
higher
than the characteristic scale that
it takes one brain area to influence another (this also demands
working close to the noise level of the measurements). A safer but
less ambitious approach, and still rather rare in the field, would
study tomographically brain activity extracted from real time
(un-averaged) MEG data filtered in the range say 3-200 Hz and
after removing the interference from the mains and its harmonics
and contributions from strong physiological sources like the eyes
and heart. Such analysis would then map brain activity at about
the timescale of transitions in the brain. The output of real-time
tomographic analysis is an attempt to describe what is happening
in the brain with minimal assumptions. Assuming that such a
reconstruction is possible, statistics on the tomographic solutions
