An Approach to Metal Artifact Reduction in CT Image via the Combination of Segmentation and Reconstruction
An algorithm for reducing metal artifacts in CT image via the combination of segmentation and reconstruction is proposed in this paper. In order to obtain the reprojection data, smoothing based on anisotropic Gaussian filter is first applied to denoise the image and smooth the streak artifacts in the CT image while preserving the important edge information as well as the shapes of the non-metallic objects around. Then we segment the metallic object regions from the non-metallic regions with maximized mutual information (MI) and replace the value of the prior regions with that of soft-tissue around to create a class termed fake tissue in the sinogram. Next, linear interpolation is applied to the fused sinogram to avoid generating new artifacts. Finally, we use computationally efficient filtered backprojection (FBP) method to reconstruct the final image without metal artifacts. It is proved, compared with conventional algorithms based on linear interpolation and model image reprojection, the proposed algorithm can reduce the metal artifacts in CT image with better performance, especially for the CT image with streak artifacts.
Anisotropk-Gaussian-filter MIMS Reconstruction Metal artifact Computed Tomography
Jiaxin Chen Ziyu Chiang
Electronic Information College Henan University of Science and Technology Xiyuan 48,471003 Luoyang, China
国际会议
重庆
英文
26-29
2011-01-21（万方平台首次上网日期，不代表论文的发表时间）