An Automatic Detection of the ROI Using Otsu Thresholding in Nonlinear Difference EIT Imaging

Abstract

Inverse problem of electrical impedance tomography is highly ill-posed therefore often prior information is used to have a satisfactory and stable solution. Recently, introduced non-linear differential imaging method estimates the initial and difference in conductivities simultaneously and is efficient in handling modeling errors. The non-linear parameterization of conductivity enables to use different regularization schemes for background and region of interest (ROI). Identifying the ROI without any prior information can be beneficial in improving the reconstruction performance. Therefore, in this paper, automatic detection of ROI is introduced using Otsu thresholding method and is then used with non-linear differential imaging. The proposed non-linear differential imaging with Otsu method (NDIWO) considers different regularization methods, i.e., total variational approach with ROI and smoothness prior with background regions during reconstruction. Numerical and experimental studies are performed to test NDIWO method for two-phase flow and thorax imaging and the performance is compared with absolute and linear difference imaging. The results indicate that the proposed NDIWO method has improved reconstruction performance compared with conventional absolute and linear difference imaging.