通过应用线性八叉树将心脏三维模型离散成体素以快速提取并显示局部任意感兴趣血管，把三维模型离散成体素后，利用多边形交互区域选择感兴趣体积内部体素。根据26-邻接体素的拓扑关系和体素互信息相似度比较结果，最终确定最佳深度和最佳互信息阈值分别为14和-1.375，以此来寻找感兴趣体积中同一组织的体素集合。最终实现了三维医学影像中任意感兴趣血管的精确显示，该技术可被视为用一个提取工具将任意局部三维模型进行提取并进行四维动态显示的过程。实验结果表明：与C-V三维分割算法精度90.1%相比，本分割算法平均精确度达到96.02%；运行时间从13.8 s缩短为10.7 s；四维播放帧数最大40 FPS,基本满足了血管三维分割的临床需求。该算法不仅可以快速地分析局部病灶的生理学特点和病理特征，而且让医生更加直观、便利地观察病人心脏任意局部血管的实际运动状况，以便做出临床决策。

In the paper, the linear octree was used to separate the 3D model of the heart into voxels and to achieve the purpose of rapid display of any interested vessels. After the 3D model is separated into voxels, the polygon area is used to select inside voxels of VOI. Based on the topological relation of 26-adjacent voxels and the similarity of voxel mutual information, the final optimal depth and best mutual information threshold are confirmed as respectively14 and-1.375, thus discovering the voxel set of the same tissue in the volume of interest. The technology ultimately achieves the precise display of any VOI from a medical image. This technique can be viewed as a process of extracting any local three-dimensional model with an extraction tool and performing a four-dimensional dynamic display. Experimental results show that compared with C-V 3D segmentation algorithm accuracy of 90.1%, the average accuracy of this segmentation algorithm has reached 96.02%；running time is shortened from 13.8 s to 10.7 s; 4D playback framer has reached its maximum 40 FPS, which basically meets the clinical needs of vascular 3D segmentation. It is not only a good way to quickly analyze the physiological and pathological characteristic for the local lesion, but it is more intuitionistic and convenient for doctors to observe the actual condition of any local blood vessel of the patient’s heart, which contributes to clinical decision-making.