Image reconstruction for electrical impedance tomography using radial basis function neural network optimized with adaptive particle swarm optimization algorithm

Image reconstruction with electrical impedance tomography (EIT) is a highly nonlinear, underdetermined and morbid inverse problem. Since traditional methods cannot achieve high accuracy and the reconstruction process is usually time-consuming, a radial basis function neural network based on adaptive particle swarm optimization (APSO-RBFNN) method is proposed and used for the image reconstruction. 15 000 simulation samples are established through numerical simulation, which are divided into the training set and test set. After network training, the image correlation coefficient (ICC) on the test set is 0.95, and the simulation results verify the effectiveness of the proposed APSO-RBFNN method. When the Gaussian white noises of 30, 40 and 50 dB are added to the test set, the ICCs are 0.90, 0.92 and 0.93, respectively, which proves the robustness of the proposed method. The reconstruction results for the samples with more targets show that the proposed method has good generalization ability. In addition, the experiment data test results of an 8-electrode EIT system show that the proposed APSO-RBFNN method has better image reconstruction results compared with the Tikhonov and RBFNN methods. © 2020, Science Press. All right reserved.