Biomedical Engineering Reference
In-Depth Information
Fig. 3.37.
Results of dipole estimation using the PT algorithm: AEF (
left
)and
SEF (
right
). The increased area of the dipole locations for SEF reflects the lower
SNR. The intensity is color coded for the SEF image from blue (lowest intensity)
to red
threshold. Measurements were recorded for five subjects with an 80-channel
magnetometer system, and were averaged from 70 runs. The noise standard
deviation ranged from 17 to 27 fT, for a 5 to 15 dB SNR. Depending on the
subjects, the source locations showed different patterns, that corresponded
to the contralateral, and sometimes the ipsilateral, first and second somato-
sensory cortices (Fig. 3.37). These first results obtained on real data, even
with a low noise ratio, are consistent with previous studies. Other practical
applications of the MCMC method are planned, both for cognitive sciences
and clinical applications.
Current Element Distribution.
The conventional procedures for analy-
zing the magnetic field source in biomagnetism are to estimate the location,
the direction, and the intensity of an equivalent current dipole which is as-
sumed to be representative of biological activity. Iso-magnetic field mapping
is mostly used to determine the location of the dipole, by finding the two
extrema of the magnetic field on the surface being observed. The dipole is
estimated directly from the extrema or calculated by some mathematical
procedure such as least square fitting of a dipole model.
To estimate a single dipole, the mathematical operation yields a unique
solution in the solution space, which consists of six coordinate corresponding
with the dipole parameters. A dipole is described by parameters of three
components of position, two of direction, and one of intensity. If we have
magnetic field signals without any noise, these unknown parameters are esti-
mated with at least six independent observations. In this case, the model is
evaluated by the value of the residue between the observed data and ma-
gnetic field generated by model. If we expand the model and adopt a more
