颅底外科手术导航的配准跟踪方法研究

Endoscopic surgery is the major method for the diagnosis and treatment of skull base tumors. But the skull base tumors are often located adjacent to the brain stem, cranial nerves and important blood vessels, and the clinical operation is difficult. Multi-modal image-guided surgical navigation technology can compensate for the missing three-dimensional information of endoscopic images, thus avoiding accidental injury to these important tissues and organs during surgery. However, due to the low accuracy of registration between multi-modal images and intraoperative instruments, it does not satisfy the requirements of clinical surgery. Therefore, the project is intended to study on the basic methods of registration and tracking involved in the skull base surgery navigation. The adversarial network is introduced to evaluate the similarity between multi-modal images, then the full convolution network is used to realize the automatic non-rigid registration of multi-modal images. The dynamic information of endoscopic sequence images is employed to reconstruct the three-dimensional structure of the nasal cavity, and then the spatial correlation between endoscopic image and preoperative 3D image is registered. The intraoperative motions of patient and surgical instrument are obtained by optical tracking, so that the intraoperative registration error can be reduced by hybrid decision of image registration and optical tracking. The related researches will be applied to improve the accuracy and stability of surgical navigation system. This work has far-reaching significance for the clinical application of skull base surgery navigation system.

内镜手术是颅底肿瘤诊疗的主要手段，但颅底肿瘤往往位置较深，毗邻脑干、颅神经及重要血管，临床手术操作难度较大。多模态影像引导的手术导航技术可弥补内镜图像所缺失的三维信息，从而避免术中误伤上述重要组织器官。但因多模态图像间的配准与术中器械跟踪精度较低，尚无法完全满足临床手术需求。为此，本项目拟开展颅底外科微创手术导航所涉及的配准跟踪基础方法研究。研究内容包括引入对抗生成网络评估多模态影像间的相似性，使用全卷积网络实现多模态影像自动弹性配准；利用内镜序列图像的动态信息，重建鼻腔三维结构，并研究鼻腔表面结构与三维影像的快速配准方法，建立内镜二维图像与术前三维影像的空间信息关联；研究患者及手术器械在术中运动的光学跟踪方法，通过图像配准与光学跟踪混合决策，实现术中配准误差实时校正。并将理论研究成果应用于颅底外科微创手术导航，提高手术的效率与安全性。本项工作对影像引导手术导航的临床应用具有深远意义。

本项目按照预定计划开展研究，围绕颅底外科微创手术导航的配准跟踪问题开展基础性研究，以突破关键技术、提升算法可靠性为目标，深入研究多模态图像引导与光学跟踪相结合的颅底外科微创手术导航方法。搭建了一套双目光学跟踪系统和增强现实颅底手术导航系统，基于三坐标测量机的相机标定和优化的灰度矩坐标计算方法，达到了0.16mm的跟踪定位精度，自主设计光学跟踪硬件结构，将光学跟踪的刷新率提升至30Hz以上。基于特征点匹配的方法实现了跟踪系统到医学图像的自动配准，在临床实验中验证目标区域的导航误差<2mm。基于重要性排序面片渲染方法实现了导航过程中内镜图像与医学图像的虚实融合，实现术中患者位姿的快速、精准校正，延展手术视野并提供直观三维结构信息，并成功应用于颅底手术临床诊疗。研发了基于距离模型的实时空间距离分级预警功能，依距离对当前手术操作进行安全分级。在完成颅底外科虚实融合导航系统的关键技术突破后，进行了产品软件框架搭建、软件功能开发、产品外观设计、产品结构设计与加工，完成了颅底外科手术导航定位系统的小批量生产并在临床实验中验证其有效性。本项目完成了项目既定的研究目标，执行期间累计发表或录用16篇SCI期刊论文、15篇会议论文，申报国家发明专利11项，其中2项已授权。

内容获取失败，请点击重试