Electrical Impedance Tomography |
 |
Electrical Impedance Tomography (EIT) is a measurement procedure based on injecting low levels of current into the body, then measuring the potential field created by volume conduction of this current through bodily tissues. From the potential field and the known position of the injected current, properties of the body tissue can be inferred through the inverse solution. Although this method has been researched for many years, it has not yet been widely applied to medical problems. A major factor in the poor adoption of EIT is that it was initially conceived as an imaging method, and x-ray and MR techniques make much better images.
Electrical Impedance Tomography research at the NeuroInformatics Center has focused on the more constrained problem of inferring subject specific electrical conductivities for head tissues. High resolution knowledge of tissue geometries can be measured by MR imaging. By assuming that tissue conductivities are homogenous within each segmented tissue type, the EIT problem may be cast as a search over the multi-dimensional space of unknown conductivities. The search employs the forward problem with chosen parameter estimates and a function that determines the error of the forward calculation with measured data from multiple current injection pairs. As the error residuals of local inverse searches are minimized, the global search determines convergence to final property estimates based on its knowledge of how well the parameter space has been sampled.
As the results of solving the forward electrical problem depend intimately on the tissue conductivities, accurate estimation of tissue conductivities is vital for successful EEG source localization. For either problem, source localization or impedance imaging, solving the inverse search usually involves the large number of runs of the forward problem. Therefore, computational methods for the forward problem, which are stable, fast and eligible for parallelization, as well as intelligent strategies and techniques for multi-parameter search, are of paramount importance.