基于一致性扩散的非迭代快速CT金属伪影去除算法

A Novel Non-iterative Metal Artifact Reduction Algorithm Using Coherence Transport in Computed Tomography

当人体中存在金属植入物时,得到的医学CT图像会产生明显的金属伪影,影响CT图像的临床诊断价值。为了在保证成像速度的前提下,有效地去除金属伪影,使用快速行进法求解程函方程,得到投影数据的修复顺序,然后设计对应的权重函数,引入结构张量构建图像的一致性流场,确定一致性强度,最后计算得到修复之后的投影值。实验证明,该算法在计算速度上接近1维插值算法,远优于基于偏微分方程的2维校正算法,在校正效果上接近2维校正方法,远优于1维插值算法。

When metallic implants exist in human bodies,there will be severe metal artifacts which can obviously affect the clinical diagnosis.To deal with metal artifacts with acceptable imaging time,a fast marching method was used to solve the Eikonal equation and the inpainting order was obtained for projection data.By designing the corresponding weighted function and introducing structure tensor to construct the coherence transport field of the sinogram which was used to determine the coherence strength,the inpainted projection data were calculated.The experiment results demonstrated that this method not only has a good metal reduction performance as two-dimensional correction methods,but also is fast close to one-dimensional interpolation methods.