Comparison of Different Quadratic Regularization for Electrical Impedance Tomography

Abstract

Electrical Impedance Tomography (EIT) is an ill-conditioned problem in which regularization is used to calculate a stable and accurate solution by incorporating some form of prior knowledge into the solution. A number of regularization terms have been proposed for EIT since they influenced the performance of EIT a lot. However, it is difficult to evaluate the performance of different regularization terms due to the influences of optimization and regularization parameters methods. This paper compares four widely used quadratic regularization methods, including Markov Random Field (MRF), Noser, Laplace and Tikhonov, with L curve and Newton- Krylov subspace method for both simulation and phantom experiments. The results demonstrate that the MRF not only generates clear images in all trials but also have moderate imaging speed, fitting the general applications, and Tikhonov approach is fastest but with strong artifacts, fitting the application with high time resolution.