Study On Electrical Impedance Tomography Algorithm Based On Neural Network

Electrical Impedance Tomography(EIT)is a non-invasive medical imaging technology,which has the advantages of simple structure,low cost,and non-radiation.It is currently a hot research topic.However,EIT reconstruction involves serious ill-conditioned inverse problems and requires high computational costs.In order to ensure the accuracy of imaging while increasing the speed of calculation,this paper combines the neural network method with the trad itional EIT algorithm,and designs an electrical impedance tomography algorithm based on neural network.Reconstruct the electrical impedance of EIT with the excellent nonlinear fitting ability of neural network.The main research contents of this paper are as follows:1.The finite element method performs EIT forward problem operations.Set target objects of different sizes,positions and impedance,and obtain their boundary voltage by solving the EIT forward problem,so as to construct the data set required for training the neural network.2.Design the EIT reconstruction model based on the Multilayer Artificial Neural Network(MANN),and establish the connection between the voltage difference and the impedance difference through the neural network.After training and parameter adjustment,the neural network fits the inverse problem of electrical impedance tomography.The neural network model thus obtained can be used for the reconstruction of EIT.3.Design an EIT post-processing algorithm based on Generative Adversarial Network with Wasserstein distance and gradient penalty(WGAN-gp).It is proposed that the electrical impedance distribution of EIT is analogous to the picture,and the electrical impedance distribution with errors and artifacts obtained by other methods is used as input.Using the confrontation training between the generator and the discriminator,the post-processing impedance distribution is closer to the real distribution,thereby reducing the artifacts of the EIT reconstruction result and optimizing the reconstruction effect.Design simulation experiments and practicality experiments to compare the reconstruction effect of multilayer neural network with single layer neural network,Newton-Raphson method(NRM)and split Bregman method(SBM)from a qualitative and quantitative perspective,and verify the effect of EIT reconstruction model based on neural network.Experimental results show that,compared with these three traditional algorithms,multilayer neural network can quickly and accurately reconstruct EIT.Similarly,based on simulation and practicality data,compare the reconstruction results of the MANN,NRM and SBM algorithms before and after the optimized operation.The post-processing algorithm can effectively reduce artifacts and improve the reconstruction effect of the size,position and impedance of the target object.