利用轨迹修复改进CT金属伪影去除算法的研究

金属伪影严重地影响CT图像的质量及其医学诊断价值。考虑到临床数据的可靠性及时效性，正弦图修复及其改进的方法已成为近年来研究的热点。典型的NMAR方法(Normalized Metal Artifact Reduction)，多数情况下能准确地恢复原始数据，并尽可能少地残留或引入新的伪影。然而当出现高密度组织（骨头等）与金属不相邻时，就会在金属与高密度组织间不可避免地引入或残留部分伪影。本文在NMAR算法的基础上，引入了一种基于正弦图中轨迹修复的金属伪影去除方法(TNMAR）。具体地，在NMAR方法中正弦图归一化后的区域修复期间，引入了高密度组织轨迹方向的平滑修复。此方法一定程度上削减了NMAR修复后金属与高密度组织间的残留伪影，且尽可能地恢复了原图的高对比度及组织细节。临床中，有很高的应用价值。 

Improvement of CT metal artifact reduction using trace repairation

Metal artifacts drastically impair the image quality in CT images and often reduce their diagnostic value. Considering the reliability and timeliness of the clinical data, the typical NMAR(Normalized MAR) algorithm is in general successful in removing metal artifacts, and contains the least residual and newly-introduced artifacts. However, in some circumstances metals are adjacent to bones of high density, that will inevitably induce the new artifacts or residual artifacts between the bones and metals. Based on NMAR, this paper introduces the Trace-based Normalized Metal Artifact Reduction (TNMAR). Specifically, during the normalized regional repair in sinogram, we introduce the bone-trace-based repair. It considerably reduced the newly introduced metal artifacts between the bones and metals, while greatly preserving the orignal high contrast and the accuracy of tissue details. In clinical medicine, it has great application value.