【作者】 翟伟明；

【导师】 贾培发；

【作者基本信息】 清华大学 ， 计算机科学与技术， 2010， 博士

【摘要】 影像引导下介入手术导航是使用多种模态的医学影像协助医生将手术器械直接穿刺到患者体内肿瘤内部进行局部治疗的计算机辅助技术,它是集人工智能、图像处理、图形学和临床技术为一体的交叉学科研究课题,对其关键技术的研究与应用可以有效地提高微创介入手术的临床精度和治疗效果,为患者提供更加有效和低创伤的肿瘤治疗手段。本文以胸腹腔介入手术规划和导航为核心展开研究,主要工作和研究成果包括:(1)提出了一种基于图割切的医学影像快速三维分割算法。该算法基于计算统一设备架构,解决了大规模体数据三维网络流图的构造和数据结构设计问题,定义了相应的边界能量函数和区域能量函数,解决了并行最大流算法中的多线程操作冲突的问题。该算法能够有效进行肝脏和肺部的半自动分割,性能相对传统方法有明显的提升。(2)提出了一种基于热场模拟的微波热消融介入手术规划方法。该方法基于生物传热学原理对微波热场进行建模,协助医生规划手术路径,设置微波参数,计算微波能量场、组织温度场和热损伤场,进而有效地模拟和显示肿瘤热消融范围。该方法为临床医生提供了可靠的术前指导,降低了手术的难度和风险。(3)提出了一种呼吸条件下胸腹腔介入手术导航的方法。该方法可以实时定位手术器械的坐标和方向,并将其和患者的三维影像融合显示,以指导医生按照术前规划路径进行精确的手术穿刺。为了降低呼吸运动对导航精度的影响,该方法对人体位姿及患者呼吸进行综合控制,并通过术中超声CT融合显示来反馈穿刺信息,降低导航误差。(4)改进并实现了基于GPU加速的快速体绘制方法。该方法通过GPU并行计算针对大规模医学影像体数据进行实时三维体绘制渲染,并通过实时局部光照计算技术提高渲染的真实感。该方法具有良好的渲染效果和较高的渲染效率,可以满足介入手术规划和导航中对医学影像可视化的实时性要求,并具有良好的人机交互效果。(5)研制了一套适用于胸腹腔介入手术治疗的计算机辅助手术导航系统。该系统在离体实验、粒子植入动物实验中均取得了良好的术中导航精度,并且在胸腹腔的微波热消融肺癌治疗中完成多例临床手术,达到良好的疗效。更多还原

【Abstract】 Image guided intervention surgery navigation is a minimally invasive cancer ther-apy which utilizes multi-model medical image for intra-operative surgery guidance, andit is a cross-discipline research topic involving artificial intelligence, medical image pro-cessing, real-time three-dimensional graphics and its application to surgery could signifi-cantly improve surgical result and reduce surgery trauma. This thesis focus on thoracoab-dominal intervention surgery planning and nagivation, with application to percutaneousmicrowave ablation, and the major work and contributes include:(1) A novel graph cat based 3D medical image segmentation algorithm is proposed.The algorithm is based on computer unified device architecture and provided a methodfor the construction of graph data structure of massive 3D medical image data. The en-ergy function for region and boundary is designed for the optimization problem, and theodd-even operation style is introduced to solve the thread con?ict problems in parallelcomputing. Experiments show that this method is well suited for live and lung seg-mentation and the performance of segmentation is significantly improved compared totraditional methods.(2) A novel preoperative surgery planning method for percutaneous microwave ab-lation is proposed. An iterative framework for necrosis field simulation and 3D necrosiszone reconstruction is introduced here, and the necrosis model is further superimposedto patient anatomy structures. Experiments have been performed on patient with hepaticand lung cancer and the results show that this method is accurate for surgery plan andcould be used as an assistant to the clinical practice.(3) An integrated solution for thoracic cavity and abdominal intervention surgerynavigation is proposed. Human body pose immobilization and dynamic monitoring ofrespiratory status is utilized to reduce the navigation error, and realtime CT/US imagefusion is used to feedback the accurate needle insertion position. Experiments shows thatthe clinical precision of navigation could be improved by this method.(4) A GPU based real-time volume rendering method is proposed and implementedto accelerate the rendering speed. Massive medical image volume data is visualized inhigh performance GPU algorithm, and local illumination computation is implemented forhigh quality rendering. The volume rendering method could well fit the requirement for real-time visualization in surgery planning and navigation.(5) An intervention surgery planning navigation system is designed and imple-mented for thoracic cavity and abdominal surgery. Experiments shows that this sys-tem could achieve good results in clinical trials through comparison of preoperative andpostoperative CT images. The development of this technique represented a significantadvance from the traditional ways for cancer therapy and significantly extends the indi-cations of minimally invasive cancer therapy technology.更多还原