Correction of the Beam Hardening Artifacts in CT Images Using Pix2pixGAN Network

Due to the noncoincidence between the mono energetic rays-based reconstruction algorithms and the multi-energy projection data, there always exit beam hardening artifacts destroying the reconstructed CT images. A correction of the beam hardening artifacts can effectively enhance the accuracy and reliability of the scan results. In this preliminary study, we propose a pix2pixGAN generative adversarial network to do this work. Compared to the traditional approaches, this method can increase the structural similarity index measure by 11.81%, improve the decrease the peak signal-to-noise ratio by 25.59% and decrease the root-mean-square error by 62.02 ^% . These results demonstrate that the pix2pixGAN generative adversarial network can simultaneously correct the artifacts and improve the recovery of details.