Initial Results of CT Volume Metal Artifacts Reduction based on 3D U-net

Abstract

Computed tomography (CT) allows rapid acquisition of information about the internal three-dimensional (3D) structure of the scanned object. However, when the scanned object contains metal implants, it will result in metal artifacts in the reconstructed images. Metal artifacts seriously interfere with the analysis and judgment of CT images. To solve the problem of metal artifacts in CT volumes, this paper introduces 3D U-net into the metal artifact reduction of CT volume data. The 3D convolution in 3D U-net can make full use of the spatial and structural information of CT volumes, and can better maintain the internal structural information based on effectively reducing the metal artifacts. The direct processing for 3D data can improve the stability and continuity of the corrected image in all dimensions. After artifact reduction for 3D data, the information of arbitrary spatial coordinate points in the image can be obtained more accurately and effectively. To verify the performance of the algorithm, experiments were conducted using medical datasets. The results show that the method in this paper can effectively suppress metal artifacts while obtaining better stability and continuity in all dimensions.