Modeling and reconstruction methods for electrical impedance tomography

In electrical impedance tomography (EIT), we inject current patterns into an object and measure the resulting voltage responses to estimate its resistivity distribution. This dissertation focuses on developing measurement methods, finite element models (FEMs) and iterative reconstruction algorithms to obtain images from measured data rather than from computer simulation.; The quality of a reconstructed image depends upon the signal-to-noise ratio (SNR) of the measured data. We injected optimal current patterns to obtain the measured voltages with maximal SNRs. We developed iterative and direct methods to obtain these optimal currents, and concluded that the Fourier-based direct method is superior to other methods. The injection of optimal current patterns requires multiple generators. We developed diagonal-based methods to produce optimal current and voltage data using one current generator. We compared the performance of a single current-generator system with that of a 32-current-generator system.; The electric current and voltage measurements are sensitive to the contact impedance, since it is large with respect to the internal tissue impedance. Most reconstruction algorithms have either ignored or simplified its effects. We developed a FEM model to accurately describe the dominant shunting and edge effects of the contact impedance. We further used compound electrodes to minimize the sensitivity of measurements to contact impedance.; Yorkey et al. used an iterative algorithm called the modified Newton-Raphson method to perform image reconstruction. The information matrix in the modified Newton-Raphson algorithm is ill-conditioned. The ill-conditioning makes the reconstruction process very sensitive to the measurement error and may produce incorrect results. We developed a regularization method to integrate the a priori information into the reconstruction algorithm, thus improving the conditioning of the information matrix and the stability of the modified Newton-Raphson algorithm. We provided a good initial guess of the contact impedance by physical measurements to avoid local minima. We showed that updating optimal current patterns after each image iteration improves image quality.; This dissertation also investigated the effects of geometric shape variation among the human body and studied the clinical application of EIT in bladder volume monitoring.