Abstract: The line filter designed by Sato can remove most of the plate- and blob-like structures in images except for background pixels. Based on the vessel enhancement principle proposed by Sato, an improved 3D vessel enhancement algorithm is presented in this paper. Aiming to remove the unnecessary information, this algorithm takes the eigenvalues of Hessian matrix of background pixels into consideration. Clinical experiment is conducted on 3D carotid artery CTA data and cerebral vessel MRA data. Experimental results indicate that the background is greatly suppressed and vessel structures become more distinct with enhanced image contrast.

Key words: vessel enhancement, Hessian matrix, line filter, vessel segmentation

摘要： Sato的滤波器可以有效地滤除3D数据中的面状结构和球状结构，但是对于背景信息滤除效果不佳。基于Sato的血管增强原理，提出了一种改进算法。新算法将背景像素的Hessian矩阵特征值也考虑到滤波器的设计中，增强了滤波器滤除背景像素的功能。基于临床颈动脉的CTA数据和脑部血管MRA数据，做了实验分析。实验结果表明，在保留原算法优点的情况下，背景内容被大幅度移除，图像对比度得到进一步提高，血管结构更加清晰可辨。

关键词: 血管增强, Hessian矩阵, 线形滤波器, 血管分割