Image Reconstruction Improvement with Optimal Driven-measurement Pattern Selection for Electrical Impedance Tomography

Driven-measurement pattern affects image reconstruction quality of Electrical Impedance Tomography (EIT). In this paper, optimal pattern selection approach based on machine learning is proposed. It converts the problem of driven-measurement pattern optimization to pattern recognition. It is simple, effective and fast. The selected optimal pattern is used for image reconstruction, which is aiming at an improvement. Support Vector Machines (SVM) is used to select the optimal pattern. Firstly, pattern candidates are generated. The optimal pattern is considered to be selected from them. Secondly, optimal pattern selection approach is introduced. SVM with estimated conductivity distribution as features is used for pattern recognition. Thirdly, the proposed approach is verified by both simulation and experiment. Average imaging correlation coefficient (ICC) of the reconstructed images for simulation models is improved 7.6163% from 0.7891 to 0.8492. In experiment, average ICC is improved 2.9439% from 0.7541 to 0.7763.