SIA OpenIR  > 机器人学研究室
基于神经信号识别的颅内电极精准植入关键技术研究
Alternative TitleThe Research on the Key Technology of Precise Intracranial Electrode Implantation based on the Identification of Neural Signal
曹蕾
Department机器人学研究室
Thesis Advisor赵忆文 ; 刘浩
Keyword颅内电极精准植入 立体定向手术机器人 电极—组织交互 柔性植入路径规划 神经信号识别靶点
Pages116页
Degree Discipline模式识别与智能系统
Degree Name博士
2020-05-29
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract神经电生理技术是一种将电极植入到特定的组织位置,并利用电生理仪器记录神经细胞电活动的技术,在揭示脑机制以及诊断治疗脑疾病等方面都有着广泛应用。神经电生理技术中最关键的步骤就是如何精准地将电极植入到最有效的靶点处。影响电极植入精准性的原因主要有以下几点:首先,通过立体定向架手工植入电极操作繁琐,在多靶点植入时容易造成额外损伤;第二,由于电极植入脑组织时,组织会产生形变,进而导致规划靶点位置发生偏移;第三,由于脑组织结构的复杂性,仅通过直线路径有时无法实现安全精准的植入;最后,由于缺少神经电生理信号反馈,仅靠术前影像无法准确识别理想的生理靶点。为了解决上述问题,本论文在地方项目的资助下,研究立体定向机器人的空间配准方法,以实现根据术前图像定位规划靶点,研究电极与脑组织的交互特性以精准定位“物理靶点”,研究柔性避障植入以规划到达物理靶点的最优路径,研究基于神经信号的自动识别方法以精准定位“生理靶点”。具体的研究内容如下:为了实现颅内电极自动植入,本研究搭建了立体定向手术机器人系统平台。提出了两种基于无框架的坐标系配准方法,其一,基于力传感反馈实现机器人的拖动控制,通过点云配准计算机器人空间与图像空间的坐标映射关系;其二,为了简化配准流程,利用光学导航系统同时实现器械标定以及坐标系配准。通过定位实验评估了坐标系配准方法的配准精度,结果表明,基于光学导航的配准方法能够快速且精准地实现机器人与图像之间的配准。最后,利用该系统平台完成了动物定位植入实验。验证了立体定向手术机器人系统辅助定位的准确性。为了弥补由于组织形变造成的物理靶点偏移,本研究以电极植入脑浅表区为例,对电极与硬脑膜交互的不同阶段进行建模。从形变、内耗以及回弹三方面描述了电极刺穿硬脑膜的交互力学特性。采用粘弹性模型对形变阶段进行模拟,得到力与组织形变的关系。基于能量守恒原理,形变阶段存储的能量由于内耗作用将损失一部分,而其余部分转换为用于组织形变恢复的能量。基于模型分析得到硬脑膜的回弹位移与穿刺力、穿刺位移以及植入速度有关。最后,搭建了电极植入的实验平台,在人工组织上进行了多组不同速度的植入实验,验证了模型的准确性并对模型参数进行了辨识。并且基于新的测试速度完成了植入实验,结果表明该模型能较好地模拟电极植入硬脑膜的整个过程。并且根据模型方程估计的回弹位移与真实测量的回弹位移基本一致。根据回弹位移,可以对电极植入的靶点位置进行校正,实现电极植入脑浅表区的精准可达。为了将电极安全地植入到校正后的物理靶点,考虑大脑组织的复杂性,利用柔性的同轴管机器人(CTR)辅助电极以曲线路径植入到靶点位置。针对不同的颅内解剖环境,不仅需要对植入路径进行规划,还需要设计满足解剖空间需求的CTR构型参数。本研究构建了一组嵌套的优化问题,首先,对于给定的CTR设计构型,考虑CTR本体形状避障,基于RRT*算法生成路径扩展树,并利用路径成本函数对路径节点进行优化。然后,根据路径成本以及CTR稳定性需求构建CTR设计成本函数,通过迭代得到CTR构型参数。最后,在模拟解剖环境中,将该方法与其他路径规划算法进行了对比,验证了算法的可用性和优越性。并得到成本最小的CTR设计参数,保证CTR能够安全精准地以曲线路径植入到物理靶点位置。利用CTR的路径通路,可将电极通过最内层管植入到靶点位置,进而实现神经信号采集等操作。为了解决医学影像难以精准定位生理靶点的问题,可通过微电极记录技术采集电极植入过程中不同位置的神经电信号活动。本研究根据神经信号特征对电极经过的脑功能区域进行分类识别,进而分析电极与脑功能区域的相对位置,校正理想的生理靶点。首先从微电极记录中提取了大量特征,并根据神经信号的类型对其进行分组。选择单一特征组作为分裂式层次聚类算法中每次分裂的输入样本。通过基于遗传算法的特征组组合选择方法,获得最优的聚类结果。最后与其他相关研究方法进行了比较,在真实的深脑电刺激手术(DBS)数据集上验证了该算法的优越性。根据算法的聚类结果,分析了目标神经核团的解剖边界以及电活动特性。为了实现在线识别生理靶点位置,本研究利用无监督随机森林(RF)算法,通过离线聚类和在线识别来识别目标神经核团。为了进一步优化算法,在算法中加入了特征选择(FS)方法,该方法包括基于轮廓系数的特征重要性排序以及基于轮盘赌选择的特征子集搜索。最后,在DBS数据集上评估了FS对无监督RF算法的优化效果。结果表明,使用FS方法可以提高在线识别的准确性并减少计算时间。脑功能区域的自动识别结果有助于实时校正理想的生理靶点位置。综上所述,本研究从影响颅内电极精准植入的主要问题出发,解决其对于物理靶点和生理靶点两方面的定位精准性难题。对立体定向机器人的空间配准方法进行研究,对电极与组织交互产生的组织形变进行分析,对复杂解剖环境设计柔性机器人实现曲线路径规划,基于神经信号反馈对神经核团靶点进行自动识别。本论文通过实验及仿真对以上研究方法进行了验证分析,研究结果为实现颅内电极精准植入提供了解决方案以及实验平台。
Other AbstractThe neuroelectrophysiological technology is a technique in which the electrode is implanted into specific region and the electrophysiological instrument is used to record the electrical activity of nerve cells. It is widely used in revealing the brain mechanism and diagnosing and treating brain diseases. The most important step in the neuroelectrophysiological technology is how to precisely implant the electrode to the most effective target. The main factors affecting the accuracy of electrode implantation can be described as follows. Firstly, the process of traditional electrode implantation based on the stereotactic frame is tedious. And it is easy to cause additional damage when there are more than one electrodes should be implanted. Secondly, when the electrode is implanted into the brain tissue, the tissue will be deformed, which will lead to the offset of the target. Thirdly, due to the complexity of the brain tissue structure, sometimes the electrode cannot be implanted safely and precisely through a straight path. Finally, due to the lack of the feedback of neuroelectrophysiological signal, the ideal physiological target cannot be localized accurately by the preoperative image. In order to solve the problems above, with the support of local project, the space registration method of stereotactic robot was studied to localize the target according to the preoperative image. And then, we studied the interaction between electrodes and brain tissue to precisely localize the “physical target”. And the optimal path to reach the physical target through flexible implantation with obstacle avoidance is studied. Finally, we studied the automatic identification method based on neural signal to precisely localize the “physiological target”. The main research contents can be described as follows. In order to achieve the automatic electrode implantation, a stereotactic surgical robot system was established in this study. Two registration methods based on frameless are proposed. Firstly, the robot was controlled to record the registration points by drag control based on force sensing. The coordinate mapping between the robot and image space can be obtained through point cloud registration. Secondly, the instrument calibration and coordinate registration are simultaneously implemented by using the optical navigation system. The accuracy of the registration method was evaluated by localization experiment. The results showed that the registration method based on the optical navigation system can achieve the registration between the robot and image quickly and precisely. Finally, the localization accuracy of the stereotactic surgical robot was verified by animal experiment. In order to compensate the offset of the physical target which is caused by tissue deformation,we take the electrode implantation into the superficial region of brain as an example. The different stages of the interaction between electrode and dura mater were modeled. We established an energy model which can be described from three parts to represent the mechanical properties of the electrode-dura interaction. The deformation part is described through a viscoelastic model and the relationship between force and tissue deformation can be obtained. Based on the conservation of energy, one part of the energy stored in the deformation stage will be lost due to internal friction. And the other part will be converted into the energy which can lead the tissue spring back. Based on the whole model, the springback displacement can be calculated by the puncture force, puncture displacement and implantation velocity. Finally, a series of experiments are implemented on the artificial tissue to validate the model and identify the character parameters. And the validation experiments in other new velocities show that the proposed model can well represent the whole process. The estimated springback displacement according to the model equation is basically consistent with the real springback displacement measured by the displacement sensor. According to the springback displacement of the dura mater, the target position can be corrected to ensure the accuracy of the electrode implantation into the superficial region of brain. In order to safely and precisely implant the electrode into the corrected physical target, considering the complexity of brain tissue, a concentrate tube robot (CTR) was designed to implant into the target through a curved path. According to different anatomical environment, not only the implantation path should be planned, but also the configuration design of CTR which can meet the requirement of anatomical space should be optimized. In this study, a nested optimization problem is constructed. Firstly, for the given CTR design, considering the CTR shape to avoid obstacles, the path tree can be generated based on the RRT* algorithm. And the path node can be optimized by using the path cost function. Then, the design cost function of CTR is constructed according to the path cost and the CTR stability requirements. And the CTR design parameters can be obtained through the iteration. Finally, the method we proposed was compared with other path planning algorithms based on the simulated anatomical environment, and the superiority of the algorithm we proposed was verified. Furthermore, the CTR design with the lowest cost is obtained to ensure that the CTR can be safely and precisely implanted into the target position through the curved path. By using the path of CTR, the electrode can be implanted into the target through the innermost tube to implement signal acquisition.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/27171
Collection机器人学研究室
Affiliation中国科学院沈阳自动化研究所
Recommended Citation
GB/T 7714
曹蕾. 基于神经信号识别的颅内电极精准植入关键技术研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2020.
Files in This Item:
File Name/Size DocType Version Access License
基于神经信号识别的颅内电极精准植入关键技(17709KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[曹蕾]'s Articles
Baidu academic
Similar articles in Baidu academic
[曹蕾]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[曹蕾]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.