Biomedical Engineering Reference
In-Depth Information
of the head (e.g. thalamus) and at the crest of gyri are close to
silent sources. The patterns of activity identified with MEG are
not very meaningful on their own because they lack anatomical
context. The background anatomy must be provided by other
methods, usually MRI, and the process of combining the back-
ground anatomy and the functional information requires consid-
erable effort to ensure accurate coregistration between the two
modalities for each subject and experiment.
5.2. Advantages
MEG is a completely non-invasive method with superior tempo-
ral resolution. With appropriate analysis methods, it can provide
accurate localization of different brain regions activated simul-
taneously. The MEG signal depends weakly on the conductivity
changes in the brain and simple models can provide accurate esti-
mates of the magnetic field generated by a source in the brain.
The insensitivity to radial sources adds to the discriminability of
MEG, especially for sources in sulci.
6. Appendix 1:
Lead Field
Analysis
The measurements
d
m
(with m labeling sensors) depend on the
primary current density vector
J
(
r
) via an integral over all regions
with primary currents (the source space, Q),
Q
ϕ
m
(
r
)
d
m
=
·
J
(
r
)d
r
(8.1)
The vector function
ϕ
m
(r) is known as the lead field and it
describes the sensitivity profile of the m
th
sensor. The lead field
is completely determined by the geometric properties of the coils
making up each sensor and the conductivity details of the bio-
logical medium. Since only vectors with non-zero overlap with at
least one lead field can contribute to the signal, we can express
the unknown current density vector
J(r)
as a linear sum over the
lead fields, modulated by some unknown function,
, which, in
its most general form, can be a function of location
and
of the
unknown
J
(
r
) itself, i.e.,
ω
J
(
r
)
=
A
m
ϕ
m
(r)
ω
(r, J(r))
(8.2)
m
The wMN solution follows from the tempting choice of setting
ω
= ω
MN
(r), i.e assuming that the function,
ω
(r, J(r))
, does not
depend on any property of
J
(
r
),
J
(
r
)
=
A
m
ϕ
m
(r)
ω
MN
(r)
(8.3)
m
