多模态小型猪心脏图像的分割与配准

发布时间：2018-04-04 15:06

本文选题：多模态　切入点：图像分割　出处：《西安电子科技大学》2015年硕士论文

【摘要】：缺血性心脏病已成为世界范围内导致死亡的首位疾病,医学影像技术是对缺血性心脏病进行准确诊断、优化治疗方案和疗效评估的重要技术手段。新型治疗方法的基础研究更是必须综合利用几种跨模态的医学成像技术,从功能和结构多个层次上实现对心血管结构和心肌功能的全面准确评估。本文利用小型猪构建在体缺血性心梗模型,采用CTA成像技术对心脏的结构信息进行高分辨率成像;采用18F-FDG进行PET/CT成像监测心肌代谢功能;采用靶向血管新生的RGD进行PET/CT成像,观测心梗后心肌的血管新生情况。但是由于成像设备之间的差异、各模态成像的时间的不同以及动物在各成像设备之间移动造成姿态的不同等原因,不同模态的影像信息无法在空间上达到匹配,为了综合性的观察结构信息与功能信息,需要进行跨模态小型猪心脏图像的分割与配准。本文针对小型猪心梗模型多模态在体影像学评估,研究了小型猪心梗模型CTA与PET/CT功能图像的配准融合问题。首先,本文实现了单模态小型猪心脏图谱的制作并规定了图谱制作的流程方法(包括CTA图谱、PET/CT结构数据图谱)。对小型猪心梗模型的18组原始数据(每组数据中包括6组CTA数据、6组FDG-PET/CT数据、6组RGD-PET/CT数据)进行了图像预处理工作,依据心脏解剖学知识,对心脏进行了人工分割,针对每种模态制作图谱库,采用基于配准的多图谱自动分割方法,利用每种模态的图谱库实现了小型猪心梗模型CTA图像与PET/CT结构图像同模态的自动分割。其次,基于心脏自动分割的结果,本文设计了CTA图像与PET/CT结构图像之间配准的方法,将CTA图像自动分割产生的CTA心脏标记图谱与PET/CT结构图像自动分割产生的CT心脏标记图谱进行配准,利用图谱图像有效的避免了不同模态间图像灰度值分布差异较大的问题,完成了小型猪心梗模型的跨模态配准工作。最后,应用跨模态配准产生的形变场信息,将PET/CT功能图像进行形变,与CTA图像进行融合,完成了结构信息与功能信息的融合。并通过体绘制将最终融合结果进行三维重构,为直观观察心梗状态与血管新生提供技术支持。本文实现了小型猪心脏同模态分割、异模态配准的完整流程,并设计了心脏分割与配准软件平台来对影像数据进行自动化处理,跨模态心脏图像的融合结果为医生评估心肌缺血与血管新生的情况提供了有效的医学影像学信息的支持。
[Abstract]:Ischemic heart disease has become the leading cause of death in the world. Medical imaging technology is an important technical means to accurately diagnose, optimize the treatment plan and evaluate the curative effect of ischemic heart disease.The basic research of the new treatment method must make comprehensive use of several kinds of cross-modal medical imaging techniques to realize the comprehensive and accurate evaluation of cardiovascular structure and myocardial function from multiple levels of function and structure.In this paper, the model of ischemic myocardial infarction in vivo was constructed in miniature pigs, and the structural information of the heart was imaged with CTA imaging technique, the myocardial metabolic function was monitored by 18F-FDG PET/CT imaging, and the PET/CT imaging was performed by targeting angiogenic RGD.Myocardial angiogenesis after myocardial infarction was observed.However, because of the difference of imaging equipment, the different imaging time of each mode and the different reasons of animals moving among the imaging devices, the image information of different modes can not be matched in space.In order to observe the information of structure and function comprehensively, the segmentation and registration of cross-modal miniature porcine heart image is needed.In this paper, the registration and fusion of CTA and PET/CT functional images in miniature porcine myocardial infarction model in vivo were studied.Firstly, the single mode miniature porcine heart map is made and the flow chart (including the CTA map and the structure map of PET / CT) is defined.Eighteen groups of original data (including 6 groups of CTA data, 6 groups of FDG-PET/CT data and 6 groups of RGD-PET/CT data) of myocardial infarction model of miniature pigs were preprocessed. According to the knowledge of heart anatomy, the heart was segmented artificially.In this paper, the automatic segmentation method of multi-atlas based on registration is adopted for each modal making atlas library, and the automatic segmentation of CTA image of miniature pig myocardial infarction model and PET/CT structure image is realized by using the atlas library of each mode.Secondly, based on the result of automatic heart segmentation, this paper designs a registration method between CTA image and PET/CT structure image.The CTA cardiac marker produced by automatic segmentation of CTA image is registered with the CT cardiac marker produced by automatic segmentation of PET/CT structure image, and the problem of large difference of gray value distribution among different modes is avoided effectively by using the map image.The cross-modal registration of miniature pig myocardial infarction model was completed.Finally, using the deformation field information generated by cross-modal registration, the PET/CT function image is deformed and fused with the CTA image, and the structure information and the function information are fused.The final fusion results were reconstructed by volume rendering to provide technical support for visual observation of myocardial infarction state and angiogenesis.In this paper, the complete flow of the same mode segmentation and different mode registration of miniature pig heart is realized, and the software platform of heart segmentation and registration is designed to automate the processing of image data.The results of intermodal cardiac image fusion provide effective medical imaging information for the assessment of myocardial ischemia and angiogenesis.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：R541;R817.4

【相似文献】