Probing depths with a two-dimensional electrical impedance imaging system

This dissertation describes the development and testing of a new image reconstruction algorithm and compensation procedures for electrode related problems in electrical impedance imaging. The algorithm and compensation measures were created as part of an effort to add depth visualization capabilities and a new electrode system to a 2D impedance imager in development at The George Washington University. The compensation procedures specifically address problems encountered with the alignment of the imager's electrode arrays and malfunctioning electrodes. Phantom studies were used to validate the compensation measures and the new visualization algorithm.;High conductivity objects were embedded in a series of agar phantoms. The new algorithm was utilized with the correction procedures to reconstruct the conductivity distributions of the phantoms. Assessments of each recovered conductivity were based on mean squared error (MSE) in the conductivity estimate and the center of mass of any high conductivity regions (in the recovered conductivity distribution) relative to the actual embedded object's location. The same metrics were also used to compare the new visualization algorithm to the two imaging methods it was derived from: Gauss-Newton and dipole basis reconstruction.;The new imaging technique when combined with the correction procedures generally achieved MSE's lower than Gauss-Newton and localized the embedded objects comparably to the dipole basis reconstruction method. In fact, for most of the examples shown, the faux tumors were correctly localized to within a few millimeters. Also image artefacts attributable to the alignment and malfunctioning electrode problems were noticeably reduced in some of the examples where the corrective measures were applied (a copy of this document and related supplemental material can be procured from: c i d - 0 6 0 4 d 2 b 6 8 d d 7 9 a e 3 [dot] o f f i c e [dot] l i v e [dot] c o m / s e l f [dot] a s p x / P u b l i c / t h e s i s [dot] p d f . Simply remove all the spaces and replace the [dot]'s by actual periods).