Algorithm Design And Optimization For Multi-frequency Electrical Impedance Tomography

Multi-frequency Electrical Impedance Tomography(mf EIT)is a new distributed parameter detection technology with advantages of non-intrusion,non-radiation,rapidity,convenience and low cost.It can be used for long-term clinical monitoring of diseases and has broad application prospects in the field of medical imaging.However,the image reconstruction of electrical tomography is an ill-posed nonlinear inverse problem,which leads to the instability of the solution process.In order to improve the accuracy of reconstructed images,the following work has been carried out: the regularization algorithms dealing with the ill-posedness,the sensitivity matrix optimization to reduce the errors in the difference imaging and the weighted frequency difference method.(1)A spatial adaptive regularization method based on L1 and L2 regularization was proposed which combines the data characteristics of ET difference imaging and the penalty characteristics.It can automatically assign the form of the regularization term according to spatial information updated from the iterative result of each step,instead of using fixed L1 or L2 norms over the whole image.The feasibility of the proposed method was verified by simulation and experiment results.The imaging results demonstrate that that the proposed method achieves superior performance to the traditional fixed L1 or L2 regularization method in preserving sharp edges of the inclusions,suppressing artifacts in background region and restoring the size of inclusions.(2)The error introduced by using the sensitivity matrix of reference field for difference imaging and the source of the error were analyzed.The relationship between the sensitivity matrix and the change of conductivity distribution is discussed and summarized.The optimization scheme of sensitivity matrix in differential imaging is presented and verified.A regularization algorithm based on sensitivity matrix optimization is proposed to improve the imaging accuracy.(3)The weighted frequency difference method was analyzed and expanded.The multiple weighted frequency difference was calculated by using the boundary voltages at more than three excitation frequencies to remove the influence of various background media when the field was composed of a variety of non-uniform high-contrast media with frequency dependence,and the imaging results were given for verification.In addition,the imaging results for each part of medium were given,respectively.