A Novel Combined Regularization Algorithm of Total Variation and Tikhonov Regularization for Open Electrical Impedance Tomography
Jinzhen Liu,Ling Lin,Weibo Zhang,Gang Li
DOI: https://doi.org/10.1088/0967-3334/34/7/823
2013-01-01
Physiological Measurement
Abstract:A Tikhonov regularization method in the inverse problem of electrical impedance tomography (EIT) often results in a smooth distribution reconstruction, with which we can barely make a clear separation between the inclusions and background. The recently popular total variation (TV) regularization method including the lagged diffusivity (LD) method can sharpen the edges, and is robust to noise in a small convergence region. Therefore, in this paper, we propose a novel regularization method combining the Tikhonov and LD regularization methods. Firstly, we clarify the implementation details of the Tikhonov, LD and combined methods in two-dimensional open EIT by performing the current injection and voltage measurement on one boundary of the imaging object. Next, we introduce a weighted parameter to the Tikhonov regularization method aiming to explore the effect of the weighted parameter on the resolution and quality of reconstruction images with the inclusion at different depths. Then, we analyze the performance of these algorithms with noisy data. Finally, we evaluate the effect of the current injection pattern on reconstruction quality and propose a modified current injection pattern. The results indicate that the combined regularization algorithm with stable convergence is able to improve the reconstruction quality with sharp contrast and more robust to noise in comparison to the Tikhonov and LD regularization methods solely. In addition, the results show that the current injection pattern with a bigger driver angle leads to a better reconstruction quality.