关键词: 金属伪影, Non-Local前置滤波, 最大互信息量分割, “伪组织”类

Abstract: To develop a corrective method in which the distorted segments in sinogram are identified and interpolated using non distorted neighbor projections，to reduce distorted tomography metal artifacts caused by high-density objects.First，the Non-Local Pre-Filter （NL-PF） reduces the noise content and smoothes streak artifacts in CT image.Next，the filtered image is segmented into several regions by Mutual Information Maximized Segmentation（MIMS）.Then the artifacts class is converted to the CT number with the surrounding material，called “artifact-tissue” class，and after that an “artifact-tissue sinogram” is produced using forward projection method.A final image is reconstructed by the filtered back-projection from appropriately combination of original sinogram and artifacts-tissue sinogram.Phantoms studies show that metal artifacts in CT image can be eliminated effectively.Furthermore，this proposed method improves the ability of organ contours detection.And，this feature can be applied to improve the performance of radiation therapy.These studies demonstrate that the proposed method can effectively reduce CT metal artifacts caused by high-density.

Key words: metal artifacts, Non-Local Pre-Filter, Mutual Information Maximized Segmentation（MIMS）, artifact-tissue class