CompanyName: Electrical Geodesics, Inc.
Abstract: Static conductivity information
of the various head tissues is vital to proper solution of the forward problem
in EEG and MEG. We assume that the tissue geometry is already known from structural
tomography, and desire to estimate the conductivities of the major regions.
These are potentially anisotropic, but purely resistive since we are only interested
in low frequencies in the EEG range. To solve this estimation problem we use
a 3D FEM of the head for the forward problem, derived from full head CT or MRI.
The inverse problem consists of an iterative algorithm that minimizes the difference
between measured and model data for multiple current injection pairs. We have
built a device to enable selection of various injection electrode pairs from
the full set of 128 electrodes in our Geodesics Sensor Net, while allowing measurement
of the induced potential field with the remaining 126 electrodes. We can switch
between several sinusoidal current levels (1 to 100 microam!
ps) and frequencies (1 to 200 Hz) to accomodate phantom or human use. The current source is battery powered and galvanically isolated from the EEG recording system. Within 30 minutes we can acquire data from all 8,000+ pairs and then extract the signal amplitudes from the ongoing EEG.
In simulation studies with spherical models, we have found that we can robustly recover the four conductivities representing the brain, CSF, skull, and scalp to high precision and accuracy in the presence of reasonable amounts of noise. We have verified the complete hardware/software system in a single physical phantom cylinder, and are now constructing a 3-layer cylinder with known conductities for additional validation. We will next apply this methodology to human recordings with a 4-sphere model and then the realistic FEM model. The latter will first be set up with 4 conductivities, then refined in stages adding skull detail followed by ventricles, gray-white matter differentiation, and finally anisotropies.