基于模板匹配的颅缝早闭手术切割方案规划方法研究
贺佳宾
2020-05-29
页数75
学位类型硕士
中文摘要

颅缝早闭是婴幼儿常见疾病,发病率约为两千分之一,患者通常表现为颅面畸形,严重影响其外貌和脑部正常发育。目前通常采用颅骨重塑外科手术对颅骨进行切割和拼接以达到扩容和美观的目的,而外科医师凭手术经验和主观判断制定手术方案,随意性大,主观性强,难以保证最佳的手术效果。因此,该手术对医生的临床经验和操作水平都提出了很大的挑战,国内只有小儿神外顶级医院才具备高质量手术的能力,导致只有极少数患者能够接受治疗。为提高手术实施效率,特别是降低手术门槛,需要进行颅骨重塑手术术式的智能化、定量化和标准化研究。其中,如何学习并利用资深医生的手术方案和设计经验来标准化颅缝早闭手术流程,以及如何定量评估手术效果,是亟待解决的技术难题。本文以常见的矢状缝早闭和冠状缝早闭为对象,研究基于模板匹配的颅缝早闭手术切割方案规划方法,完成的主要工作如下:
1.提出了一种基于深度学习的颅骨切割方案提取方法
为学习利用并标准化资深医生手术经验,本文提出一种结合深度学习和点云处理技术的颅骨切割方案提取方法,该方法将深度学习应用于颅骨外表面的目标区域检测和实例分割问题,结合深度相机和点云处理技术实现颅骨切割方案的自动提取和数字化。实验证明该方法可以准确高效地检测出颅骨切割方案并将切割方案轨迹坐标进行三维映射,精度方面可以保证切割方案轨迹点定位误差小于2mm。
2.提出了一种理想颅骨模型生成方法
目前的颅骨重塑外科手术缺乏因人而异的理想颅骨模型来指导手术方案设计并进行手术效果评价。针对这一难题,本文首次提出了一种基于点云配准技术和点云镜像操作的冠状缝早闭患者理想颅骨模型生成方法,辅助医生在术前设计手术方案并定量评估预期术后效果。还提出了一种利用乳突和顶结节等解剖结构特征进行配准检索的矢状缝早闭患者理想颅骨模型生成方法,可以实现在正常婴幼儿颅骨数据库内的最优匹配。
3.提出了一种基于模板匹配的手术切割方案生成方法
通过前期对资深医生颅缝早闭手术切割方案的学习,本文分别建立了针对矢状缝早闭和冠状缝早闭的手术方案模板库,并提出了一种基于点云配准和轨迹投影的模板匹配方法,通过比较测试病例颅骨点云与相应模板库中模板点云之间的相似度,选择匹配程度最高的模板手术方案作为测试病例的初步手术方案,然后通过轨迹点投影快速生成测试病例的手术方案。
基于以上提出的方法,我们设计并实现了颅缝早闭手术切割方案规划软件系统,将上述功能进行模块化集成。医生使用该系统可以在现有数据的基础上扩充手术方案模板库,并利用模板库实现患者手术方案的快速生成。此外,还可以使用理想颅骨模型辅助手术方案的进一步优化设计并对术后效果进行定量评估。本文工作可以有效缩短术前方案设计时间和术中操作时间,有助于颅缝早闭手术切割方案制定的规范化和标准化。
 

英文摘要

Craniosynostosis is a common disease among infants and young children, there are about one in two thousand infants suffered from it. Patients usually present with craniofacial deformities, which seriously affects their appearance and normal brain development. Nowadays, surgery is usually taken to cut and splice the skull on the purpose of expanding and bringing the skull back to normal. However, doctors conduct operations mainly based on their experience and subjective judgment, which is arbitrary, subjective and no guarantee to the best postoperative results. Therefore, it poses a great challenge to the clinical experience and operation level of the doctors. In China, only the top pediatric hospitals have the ability to perform high-quality surgery, causing only a few infants have chance to be treated. In order to improve the efficiency of surgical operations, especially to reduce the difficulty of the surgery, it is necessary to conduct intelligent, quantitative and standardized research on the operation method for craniosynostosis. Therefore, the urgent technical problem to be solved is how to extract and learn the surgical experience of senior doctors' to standardize the surgical plan procedure for craniosynostosis and how to evaluate the postoperative effect quantitatively. This thesis focus on the sagittal suture closure and coronary suture closure, and proposed a surgery planning method based on template matching for craniosynostosis. The main tasks completed are as follows:
1.A method for extracting skull cutting plan based on deep learning is proposed
In order to learn and standardize the experience of senior doctors’ in surgical scheme design, this thesis proposed a method combining deep learning and point cloud processing technology to extract skull cutting plan. This method applies deep learning to the instance segmentation of the skull surface. Based on detection and segmentation result of the target area on the surface, point cloud processing technology is used to realize the automatic extraction and digitization of the skull cutting trajectory. Experiments show that this method can accurately and efficiently detect the skull cutting trajectory and map the trajectory coordinates in three dimensions. Besides, this method has an accuracy of less than 2mm when localizing cutting trajectory points.
2.An ideal skull model generation method is proposed
The current surgical procedure calls for personalized and ideal skull model to guide the surgical plan and evaluate the postoperative effect. To solve this problem, this thesis proposed an ideal skull model generation method based on point cloud registration and point cloud mirroring technology for the coronary suture closure, which can assist doctors to design preoperative surgical plan and quantitatively evaluate expected postoperative effect. Besides, this thesis also proposed an ideal skull model retrieval method based on anatomical features of mastoid and apical nodules and point cloud registration for the sagittal suture closure, which can achieve optimal matching in the template library of normal infants’ skull.
3.A surgery plan generation method based on template matching is proposed
After the previous learning of cutting schemes for craniosynostosis of experienced doctors, this thesis established a template library of surgical schemes for sagittal suture closure and coronary suture closure respectively and proposed a template matching method based on point cloud registration and trajectory projection. In the experients, by comparing the similarity between the skull point cloud of the test case and the template point clouds in the library one by one, the template which has the minimal registration error is selected as the initial surgical scheme for the test case, and then generates a personalized cutting plan for the test case using the trajectory projection.
Finally, based on the methods proposed above, we have designed and implemented a craniosynostosis surgical scheme planning software system, which integrates all above functions in a modular manner. Using this system, doctors can establish and expand the surgical plan template library based on existing clinical data, and then use the template matching method to generate the test case’s cutting plan. 
In addition, doctors can use the ideal skull model to assist the further optimization of the surgical plan and quantitatively evaluate postoperative effect. In general, the method proposed in this thesis can effectively shorten the preoperative plan design time, accelerate the intraoperative operation, and help to standardize the surgery plan for craniosynostosis.
 

关键词颅缝早闭 模板匹配 深度学习 点云配准 理想颅骨模型
语种中文
文献类型学位论文
条目标识符http://ir.ia.ac.cn/handle/173211/39091
专题智能制造技术与系统研究中心_智能机器人
推荐引用方式
GB/T 7714
贺佳宾. 基于模板匹配的颅缝早闭手术切割方案规划方法研究[D]. 中国科学院大学. 中国科学院大学,2020.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
基于模板匹配的颅缝早闭手术切割方案规划方(2777KB)学位论文 开放获取CC BY-NC-SA
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[贺佳宾]的文章
百度学术
百度学术中相似的文章
[贺佳宾]的文章
必应学术
必应学术中相似的文章
[贺佳宾]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。