Electrical impedance tomography ( EIT) is a kind of imaging technology to realize the image reconstruction of electric conductivity distribution in the practical field. Traditional electrical impedance imaging algorithms have the problem of low imaging accuracy. To address this issue, a new electrical impedance image reconstruction method based on the U 2 -Net deep learning model is proposed in this paper. First, based on the U 2 -Net model, this paper innovatively proposes the concept of concatenate (CAT) for data extension, which makes the input layer of U 2 -Net simple in structure and fast in operation speed. Secondly, the simulation data set is used to train the network, and the validation set is used to select the optimal model parameters. Experimental results show that the proposed algorithm has high measurement accuracy and good robustness. This method performs better than other algorithms in the simulation data set. Finally, a new EIT imaging quality evaluation index is proposed to evaluate the performance of the algorithm, which is named as center and area error (CAE). Experimental results show that the CAE of the proposed algorithm is 4. 975, which is more accurate for the prediction of the center and area of the target object. And the imaging effectiveness is better than other comparison algorithms.