医学体数据中面向数值分析的等值面重建算法

Novel Iso-surface Reconstruction Algorithm for Numerical Analysis from Medical Volume Data

Marching Cubes是医学体数据可视化的经典算法,但生成的等值面网格存在拓扑二义性和单元质量两方面的缺陷,无法为生物组织物理仿真中的数值分析提供良好的几何模型,为此,提出一种基于数据点偏移的改进Marching Cubes算法。算法将数据场分解为点、边、面和体素四类元素;以33种剖分模式为依据,构建二义性检测索引表,通过提出的基于面状态的渐近线判别法,以统一的方式解决面二义性和体二义性问题;分析单个体素中产生退化三角形的原因,提出基于局部判别法的数据点偏移策略,使体素的活跃边与等值面近似垂直,达到提高网格质量的目的。对比实验表明,该方法在有效保证网格拓扑的基础上,显著提高了单元质量,生成的模型不仅适用于体数据可视化,还适用于进一步的数值分析。

For the purpose of numerical analysis, an improved Marching cubes algorithm is presented to solve the ambiguity and generate mesh with good quality. Firstly, the medical volume data was decomposed to the topological structure of points, edges, faces and Voxels. Secondly, ambiguity-detection index tables were constructed based on 33 ＇cases. Asymptotic decider based on face state was then proposed to solve the two classes of the ambiguity. Lastly, according to the analysis of degenerate triangle generated in a single Voxel, a novel improved strategy called data offset was proposed based on local judgment to improve the mesh quality. Compared with 33＇cases marching cubes and some other improved algorithms, this produces better results and can be used for further numerical analysis as well as visualization.