【作者】 刘志强；

【导师】 刘守斌；

【作者基本信息】 哈尔滨工业大学 ， 机械工程（专业学位）， 2018， 硕士

【摘要】 截至2015年,脑血管疾病患病率高达12‰,已严重威胁到我国居民健康,脑血管三维模型的可视化及数值模拟,可为研究学者和临床医生提供重要的诊断和治疗信息。点云数据处理是目前研究的热点,已经应用在VR、自动驾驶等新型技术中,本文将其应用到脑血管领域中。一些未经过分割处理的血管点云数据,难以得到高质量的脑血管模型和进行下一步血流分析,本文拟对具有复杂拓扑结构下的脑血管分割、骨架线提取、光滑表面重建和血流动力学分析展开研究。针对脑血管点云数据的分割,提出了基于连通性的分割方法。基于连通性的分割算法首先建立覆盖集,对覆盖集进行特征分析,制定覆盖集连通性判断标准,然后确定分割起始点,通过对覆盖集的连通性判断进行点集扩充,当试探集不连通时,则可能存在分叉区域,这时需要根部反向扩充来优化判断,最后对未处理的覆盖集继续进行扩充,直到所有覆盖集被处理,完成脑血管的局部分割。在分割的基础上,提出了基于空间中值的骨架提取方法。基于空间中值的方法,通过定义局部半径引入排斥力和稠密度权重函数,实现对骨架点的提取;然后将提取的骨架点连接成线,分别对分支上的骨架线连接和分叉处的骨架线连接提出了解决方案;最后完成脑血管骨架点的提取和骨架线的连接。脑血管表面重建是在骨架线的基础上进行的。首先对骨架线进行分层编码,重组成各个单条分支;再通过相邻等间距采样的骨架点计算中间骨架的切向量和最优截面直径,构造正交截面圆,将轮廓上的点进行三角拼接重构各单分支的表面;然后提出类Delaunay三角剖分方法处理分叉区域的表面,通过将各分支之间的“互吃”关系映射到二维平面,构造类Delaunay三角,再对空洞区域进行填补,完成分叉区域的表面过渡拼接;最后对重构的表面进行Loop细分处理,获得一个光滑准确的脑血管三维表面模型。从脑血管三维建模中得到相关血管形状与几何参数对血流动力学的研究至关重要,有助于脑血管疾病的临床研究和诊断。通过对关注区域的脑血管局部模型和狭窄情况下进行网格划分、血液流体参数设置和相关边界条件设置,进行了一个心动周期的数值模拟,分别分析了正常情况下和狭窄情况下不同时刻的血流速度、壁面压力和壁面切应力三个动力学参数。更多还原

【Abstract】 As of 2015,the prevalence of cerebrovascular diseases is as high as 15%,which has seriously threatened the health and cerebrovascular three-dimensional visualization and numerical simulation of Chinese residents and provided important diagnostic and therapeutic information for researchers and clinicians.Point cloud data processing is the current research focus,has been used in VR,autopilot and other new technologies,this article will be applied to the field of cerebrovascular.Some undifferentiated vascular point cloud data,it is difficult to get high-quality cerebral vascular model and the next blood flow analysis,this article intended to have a complex topology of cerebrovascular segmentation,skeleton line extraction,smooth surface reconstruction and blood flow Kinetic analysis to study.For the segmentation of cerebrovascular point cloud data,a segmentation method based on connectivity is proposed.The segmentation algorithm based on connectivity establishes the coverage set,analyzes the characteristics of the coverage set,formulates the judgment criteria of the coverage set connectivity,and then determines the starting point of segmentation,and expands the point set by judging the connectivity of the coverage set.When the heuristic set is disconnected Then there may be bifurcation region.In this case,we need to reversely expand the root to optimize the judgment.Finally,the unprocessed coverage set continues to be expanded until all the coverage sets are processed to complete the local segmentation of cerebrovascular.On the basis of segmentation,a skeleton extraction method based on spatial median is proposed.Based on the method of spatial median,the local radius is introduced to introduce the repulsive force and the density weight function to extract the skeleton points.Then the extracted skeleton points are connected into lines,and the skeletons at the branches are connected and bifurcated Skeletal line connection presents a solution.Finally,this thesis completes the extraction of cerebral vascular skeleton points and the connection of skeleton lines.Cerebral vascular surface reconstruction is based on the skeleton line.First,we stratified coding of the skeleton line and reorganized into each single branch.And then through the adjacent equally spaced sampling skeleton points calculate the tangent vector and the optimal section cross-sectional diameter,orthogonal cross-section circle,the outline of the points on the triangle Then the Delaunay triangulation method is proposed to deal with the surface of the bifurcation area.The Delaunay-like Delaunay triangulation is constructed by mapping the "eating each other" between the branches to the two-dimensional plane,and then the surface of the bifurcation area is spliced.Finally,the surface of the reconstructed surface is subdivided into Loop segments to obtain a smooth and accurate three-dimensional surface model of cerebrovascular.From the three-dimensional modeling of cerebral blood vessels to get the relevant vascular shape and geometric parameters of hemodynamic study is essential to help clinical research and diagnosis of cerebrovascular disease.A cardiac cycle was numerically simulated by meshing blood flow parameters and setting the relevant boundary conditions in the local cerebrovascular region of interest and the stenosis,and the effects of different time points under normal conditions and at narrow time Blood flow velocity,wall pressure and wall shear stress three dynamic parameters.更多还原