Landweber algorithm is limited to further applications due to its problems of semi-convergence and slow reconstruction speed. To solve the above issues, this paper firstly analyzes the causes of semi-convergence characteristic of Landweber algorithm from the perspective of the negative sensitivity field. Secondly, a method of data screening based on contribution degree analysis is proposed to weaken the influence of negative sensitivity fields on the semi-convergence characteristic of the algorithm. Then the valid capacitance data are selected from the original capacitance data based on the method. Finally, the reconstructed results of Landweber algorithm with the valid capacitance data and original capacitance data are evaluated by taking correlation coefficient and computation time as evaluation criteria. The results indicate that the new method effectively suppresses the semi-convergence characteristic of the algorithm, improves the convergence effect of the algorithm, and increases the image reconstruction quality and speed.