Research On Electrical Impedance Tomography Method Based On Deep Learning

Electrical impedance technology(EIT),an imaging technique that uses boundary measurement voltage data to reconstruct the conductivity distribution of an object,has been developed,tested,and referenced in recent years.Electrical impedance imaging is a non-invasive,non-invasive,nondestructive and convenient detection method.EIT is used by attaching n electrodes to the surface of the object,and then applying an excitation current safely below the human tolerance range,and reconstructing the conductivity distribution inside the object by measuring the boundary voltage.Although EIT has shown broad potential in various applications such as medical imaging and industrial process monitoring,its practical application is still limited due to the unsatisfactory quality and slow speed of traditional imaging.Aiming at these problems,an EIT imaging algorithm based on deep learning is proposed,which aims to improve the imaging quality and improve the imaging speed.The effectiveness of the proposed method is verified by simulation imaging experiments and physical imaging experiments.The main research contents of this paper are as follows:1.Generation of datasets.In this study,the finite element method was used to simulate the EIT problem.Set the target objects of different sizes and positions,obtain their boundary voltages by solving the EIT positive problem,and build the dataset required for training the neural network.The use of finite element methods and the construction of diverse datasets are crucial to the success of the proposed deep learning-based EIT imaging algorithm.2.An EIT reconstruction model based on the improved sparrow search algorithm-BP neural network is designed.By establishing a connection between voltage and resistivity through this network,the model is able to fit the inverse problem of EIT.After training and parameter adjustment,the network can accurately reconstruct EIT images,and the resulting network model provides an effective solution to the EIT reconstruction problem.3.Design an EIT post-processing algorithm based on denoising autoencoder.First,the split Bregman method is used to roughly reconstruct the image,and then the denoising autoencoder is used to denoise the roughly reconstructed image.Simulation and measured results show that for circular targets,the proposed method can achieve high accuracy artifact removal and accurate shape reconstruction.