Metal Artifact Reduction Algorithm In CT Images Based On Sinogram

When CT scans are performed on patients with implanted metal prostheses,metal artifacts may appear in the resulting CT images,which interfere with subsequent disease diagnosis and treatment,requiring reduction of the artifact images(Metal Artifact Reduction,MAR).The restoration method based on sinusoidal projection image first split from the original image to get the metal image,and then Radon transform is used to obtain the original projection and the metal projection respectively.The metal trace mask is calculated from the latter,according to which the damaged data in the corresponding area of the original projection are repaired and reconstructed to obtain the artifact removal image.In order to better repair the metal trace part,an "artifact free" image similar to the original image can be constructed,that is,a prior image,which can be used as a reference during the repair.The quality of the prior image determines the effect of the final artifact restoration,which should be as close as possible to the real image without metal artifacts.In order to obtain a good prior image,a Complementary Metal Artifact Reduction(CMAR)strategy was proposed.First,a simple interpolation method was used for rough restoration of the original image.For strong artifacts,rough restoration was used.In order to avoid tissue loss and secondary artifacts caused by rough restoration,data from the original artifact image was used,and then the residual artifacts of the reconstructed image were removed by bilateral filtering,while the edge information of the high contrast region was retained to obtain the prior image.Finally,the metal track restoration was completed based on the prior image.The metal artifact images provided by United Imaging Company were used for experiments,and the structural similarity(SSIM)and peak signal-to-noise ratio(PSNR)were evaluated.Compared with the previous method,CMAR showed better effect in artifact removal,secondary artifact suppression and tissue information retention,and the effect of artifact removal meets the clinical application requirements.At the same time,it can provide training data for deep learning method.