基于异质多区域CT扫描数据处理的表面网格自适应重建方法研究

The finite element analysis of patient-specific model often requires the analysis of multiple heterogeneous tissues that are in contact with each other as an assembly. Whether the precise reconstruction of the surface mesh from the heterogeneous multi-region CT scan data directly affects the finite element analysis results. The existing CT scanning data processing methods and surface mesh reconstruction technology, there will be poor separation accuracy, normal vector estimation is not accurate, in the heterogeneous multi-region assembly interface to produce gaps and overlap, and the emergence of dense mesh poor quality problems. The project focuses on the above issues, focusing on research: 1) the multi-phase level set CT image segmentation method of fuzzy clustering is studied, the mapping mechanism between the contours of the spatial contour points and the voxels of the two-dimensional contour pixels is constructed. The normal vector estimation of the contour point cloud data is carried out by the gradient value of the CT voxels; 2) Study on multi-phase implicit surface reconstruction theory, discussion and analysis of assembly rules for heterogeneous multi-region structure, the establishment of a heterogeneous multi- region implicit surface design rules, the convex assembly and adjacent assembly into non-manifold assembly; 3) the introduction of dynamic particle system theory to extract multi phase implicit surface, the adaptive distribution mechanism of particles is studied, and the multi-region mesh model is reconstructed according to particle density. This project provides theoretical and technical support for the digital modeling of heterogeneous multi-region structures and the subsequent finite element meshing.

个性化医学有限元分析经常需要将相互接触的多个异质组织作为一个装配体进行分析，能否从异质多区域CT扫描数据中精准重建表面网格模型，对有限元分析效果影响很大。现有CT扫描数据处理方法和表面网格重建技术，存在分割精度差、法矢估算不准确、异质多区域的装配接触面产生间隙和重叠、网格稠密质量差等缺点。本项目针对上述问题，重点研究：1）基于模糊聚类的多相水平集分割多目标图像，构建空间轮廓点坐标和二维轮廓像素点所在体素之间的映射机制，通过CT体素的梯度值进行轮廓点云的法矢估算；2）分析异质多区域结构的装配规律，建立多区域隐式曲面设计规则，将外凸装配和相邻装配转化为非流形装配，重构多相隐式曲面；3）引入动态粒子系统理论，提取多相隐式曲面的采样点，揭示粒子自适应分布机理，根据粒子分布状态，自适应重建异质多区域表面网格模型。本项目为异质多区域结构的数字化建模和后续的有限元体网格划分提供理论依据和技术支持。

个性化医学有限元分析经常需要将相互接触的多个异质组织作为一个装配体进行分析，能否从异质多区域CT扫描数据中精准重建表面网格模型，对有限元分析效果影响很大。现有CT扫描数据处理方法和表面网格重建技术，存在分割精度差、法矢估算不准确、异质多区域的装配接触面产生间隙和重叠、网格稠密质量差等问题。针对上述问题，本项目对基于异质多区域CT扫描数据处理的表面网格自适应重建方法进行研究，主要研究内容如下：（1）通过模糊聚类的多相水平集分割多目标图像，依次提取出CT图像序列中的异质多区域结构轮廓单像素边缘。建立了空间轮廓点坐标和二维轮廓像素点所在的体素之间的映射规则，根据CT体素的梯度值进行点云数据的法矢估算及重定向，成功将形成的多层二维轮廓数据集转化为带有法矢的三维点云数据。（2）提出多相隐式曲面重建技术，探明了多区域交界处出现间隙和重叠的原因，建立异质多区域隐式曲面设计规则，将外凸装配和相邻装配转化为非流形装配，采用正交多项式基函数有效抑制了交界曲面产生伪面片，对局部拟合函数进行加权叠加依次构成不同的隐式曲面，完成异质多区域表面重建。（3）研究了一种自适应表面网格重建方法，用自适应球体覆盖轮廓点，将球体中的辅助点连接起来进行自适应重建三角形网格；通过对异质多区域隐式曲面自适应粒子采样方法，使采样粒子根据多相隐式曲面的高斯曲率自适应分布，引入Delaunay三角剖分对粒子集进行三角化，完成三角网格模型重建。本项目有效解决了异质多区域网格建模过程中的一系列难题，为个性化医学有限元分析提供了高品质的表面网格模型，在临床治疗中有很大的应用前景，具有重要的理论意义和工程应用价值。