Study On Data Fusion And Regularization Algorithms Of Electrical Impedance Tomography

Electrical Impedance Tomography(EIT)is a non-invasive visualization technique,with advantages of high time resolution,non-radiation,non-invasive,low cost,and portability.Currently,it has been used in the medical monitoring,industrial testing and other fields.However,the low spatial resolution limits its development in the application fields.This paper focuses on the reconstruction algorithm of EIT,mainly including information fusion,regularization algorithm,etc.The details are as follows:1.To solve the limitation of EIT single excitation pattern,the research on the data fusion of multiple excitation patterns was carried out,and a Choquet integral-based fusion of the multiple exciting patterns method was proposed.This method could reveal the interaction of measurements between multiple patterns,then remove the redundant information,and finally selected the optimal measurements from each pattern and combined them to reconstruct image.The results showed that the proposed method can integrate the advantage of each pattern and improve the quality of reconstruction image.2.To improve the EIT reconstruction image quality of human lung,the structure and conductivity distribution of the thoracic cavity were used as priori information and added into the reconstruction algorithm.First,the impedance spectrum measurement was performed on normal and cancerous tissues of lung cancer patients,then the frequency pairs that best reflected the difference between normal and cancerous tissues were obtained through analysis,providing a priori basis for the frequency difference EIT research used for lung cancer detection.Second,an extended Kalman filtering algorithm with prior information was proposed.The algorithm utilized the organ structure and conductivity distribution information within the chest to construct regularization term,and then added it to the objective function of the extended Kalman filter algorithm,which could weaken the“ill-posed”of the EIT inverse problem.The results showed that the proposed method can effectively improve the quality of reconstruction images,and it has good anti-noise performance.Third,for the detection of lung cancer tissue and lung reconstruction of EIT,two methods for determining regularization parameter based on priori information were proposed,and then used into regularization algorithms to reconstruct image.The results showed that,compared with the classical L-curve method and the generalized cross-validation method,the two proposed methods can determine better regularization parameter,improve the quality of the reconstruction image,and have good robustness to the disturbance of prior information.3.An adaptive re-weighted L_pregularization method was proposed.The method could quickly solve the objective function of L_pregularization,and determine each value of p corresponding to each unknown variable adaptively in iterations.In addition,a convergence analysis of the proposed method in theory was also did.The results showed that compared with the classic regularization algorithms,the proposed method could improve the quality of reconstruction image with high computational efficiency and strong robustness to the disturbance of regularization parameter.4.A method to determine the regularization parameter based on a random matrix clustering was proposed.The basic properties and theoretical basis of the method were analyzed in depth,then the method was used into Tikhonov regularization algorithm to reconstruct image.The results showed that,compared with the classical L-curve method and generalized cross validation method,the algorithm could determine better regularization parameter and improve the quality of reconstruction image.