基于顶帽运算和各向异性扩散的CT图像金属伪影去除

Metal artifact reduction from CT images based on top hat operation and anisotropic diffusion

金属伪影的存在严重降低了CT图像的质量和诊断价值.现有的基于先验的金属伪影去除方法大多数都是从插值图像中获得先验图像,而错误的金属分割常常会导致最终校正图像中的伪影残留和组织结构丢失.针对这一问题,提出了一种基于顶帽运算和各向异性扩散滤波的归一化金属伪影去除算法.顶帽运算可以恢复由线性插值引起的骨骼结构丢失问题,且各向异性扩散滤波具有保留图像边缘细节同时减少噪声的作用,可以去除图像中的残留伪影,从而获得质量更好的先验图像.实验结果表明,与经典的金属伪影校正算法相比,提出的算法可以有效减少金属伪影,避免引入二次伪影,同时保留金属植入物周围的骨骼和精细结构.

The presence of metal artifacts seriously degrades the quality and diagnostic value of CT images.Most existing a priori-based metal artifact reduction methods obtain a priori images from interpolated images,and incorrect metal segmentation often leads to artifact residue and tissue structure loss in the final corrected images.To address this problem,a normalized metal artifact reduction algorithm based on top-hat operation and anisotropic diffusion filtering is proposed.The top-hat operation can recover the loss of skeletal structure caused by linear interpolation,and the anisotropic diffusion filtering has the effect of preserving the image edge details while reducing noise,which can remove the residual artifacts in the image and thus obtain a better quality a priori image.The experimental results show that the proposed algorithm can effectively reduce metal artifacts and avoid introducing secondary artifacts,while preserving the bone and fine structures around the metal implants compared with the classical metal artifact correction algorithm.