脑血管兴趣区域提取关键技术研究

The brain vessel VOI(volume of interest) extraction is the foundation of the noninvasive morphology detection and electronic pathology analysis in this organ. And it is significant in detecting and protecting brain desease.The shortcomings of the existing methods are over-reliance on the experts, support of special equipments, inefficiencies and so on. This project studies the representation model, 3D reconstruction ,segmentation of 3D model and the segmentation algorithm evaluation of the brain vessel to meet the urgent demand of the brain vessel analysis. The key technologies including :①the logical tree structure of the brain vesse is analyzed, and the brain vessel representation model is reprensented based on the Ball-B Splines which is first put forward by our research group. ② The curve skeletons extraction method based on the wave fronts of different energies is proposed and Combining the definition of Ball B-Spline Curve and envelope theory, a method of minimizing energy function is derived to compute corresponding radii of skeleton points.The interpolation and approximation method is used to realize the brain vessel reconstruction based on the Ball-B Splines. ③ The constrained random walk is put forward to realize the interactive sketch-based segmentation. And the fast geodesic curvature flow is used to optimize the cutting contour ④ The geometrical characteristics and user experience are combined to achieve the algorithm evaluation ⑤ Intergrate all these study results and realize the 3D brain vessel VOI(Volume Of Interest) extration platform. This research can not support the automatically detect and protect brain disease, but also startup the new direction in the brain Clinic Radiology research. And the system can be used in assisting cerebral aneurysm screening and diagnosing, multimodal image registration, anatomical atlas creating, interventional neuroradiology, and 3D visualization. There are a lot of need in medical examination and hygiene of the patient every year, so the application of the research must have extensive social benefits and great economic efficiency in practical use and great significance to science.

脑血管兴趣区域提取是脑血管形态无创检测和电子病理分析的基础，对脑病的检测和防治具有重要意义。现有方法存在过度专家依赖，须专用设备支持和效率低下等问题。本项目面向脑血管分析需求，研究脑血管兴趣区域提取中模型表示、重构方法、三维模型分割及算法评价等基础问题。①分析血管树状逻辑结构，根据项目组的球B样条成果，提出脑血管球B样条的表示模型。②形成基于波传播的骨架线提取算法，提出基于最小包络球半径函数模型，通过插值和逼近的方法实现脑血管球B样条重构。③研究基于约束随机行走的脑血管模型草图交互分割算法，构造快速测地线曲率流优化分割曲线。④融合几何特征和用户体验，实现分割算法性能评价。⑤集成研究成果，构建脑血管兴趣区域提取原型系统。相关研究不仅可为脑血管疾病的诊断与预防提供技术支持，同时为脑临床放射研究开辟新领域。此技术可推广到血管探测、多模式图像配准和三维可视化等相关领域，应用前景广阔，科学意义重大。

本项目面向脑血管模型分割问题，构建了交互与自动系列分割方法预评价方法。提出基于提出脑血管Willis环的交互提取算法研究，创新性地将图割法引入到模型交互分割工作中，融合模型的法向夹角与血管的形状直径函数，通过能量最小化，应用最大流量最小割算法实现Willis环准确提取。构造基于监督学习的脑血管分割方法，提出一种新的层叠泛化算法的几何分类后，通过有限隐马尔可夫模型，保证了标记结果的拓扑正确，并提高标记正确率。实现融合专家知识的脑血管分割交互验证方法，通过误差反馈与迭代学习后，研究所提出的方法可以自动处理Willis环血管缺失或增加的情况。融合本项目研究成果，设计实现了脑血管诊疗的物联网E-health平台层次结构，实现脑血管虚拟内窥图形工作站构建，已经在两家医院实现小样本病例实验，取得了良好得医生评价。同时设计应用相关方法实现了血管修复，自动配准与模型分析等系列应用。研究取得了系列成果，本项目面向脑血管模型分割问题，构建了交互与自动系列分割方法预评价方法。提出基于提出脑血管Willis环的交互提取算法研究，创新性地将图割法引入到模型交互分割工作中，融合模型的法向夹角与血管的形状直径函数，通过能量最小化，应用最大流量最小割算法实现Willis环准确提取。构造基于监督学习的脑血管分割方法，提出一种新的层叠泛化算法的几何分类后，通过有限隐马尔可夫模型，保证了标记结果的拓扑正确，并提高标记正确率。实现融合专家知识的脑血管分割交互验证方法，通过误差反馈与迭代学习后，研究所提出的方法可以自动处理Willis环血管缺失或增加的情况。融合本项目研究成果，设计实现了脑血管诊疗的物联网E-health平台层次结构，实现脑血管虚拟内窥图形工作站构建，已经在两家医院实现小样本病例实验，取得了良好得医生评价。同时设计应用相关方法实现了血管修复，自动配准与模型分析等系列应用。研究取得了系列成果，在国内、国际上发表有影响的高质量学术论文及会议论文28篇，其中发表期刊论文19篇，其中SCI论文7篇，EI论文24篇；发表会议论文8篇；申请专利8项，软件著作权6项；获得北京市科学技术奖一项；教育部科技进步奖一项，中国计算机学会科技进步奖一项，参加高水平国际会议17次，超额完成了项目研究内容。