在神经导航中，连续自动调节术前MRI与术中视频数据，以考虑脑部的变形情况

This PhD project addresses the challenge of brain shift during tumor resection by developing a wearable stereoscopic vision system. It aims to track brain deformations, register intraoperative observations to preoperative MRI, and reconstruct resection cavities. Candidates are expected to have knowledge in computer vision, image processing, and interest in algorithm development.

Job description: Project description: During brain tumour resection, deformations caused by craniotomy and surgery, known as brain shift, reduce the accuracy of neuronavigation based on preoperative MRI. These deformations can exceed one centimetre and compromise the correspondence between preoperative anatomical landmarks and the surgical scene. This PhD aims to develop a wearable stereoscopic vision system and computer vision methods to track these deformations, register intraoperative observations to MRI, and reconstruct the resection cavity. The first research strand will focus on a system comprising two RGB cameras mounted on the neurosurgeon's head. Recording from the surgeon's viewpoint is intended to reduce occlusions encountered with fixed cameras and maintain temporal continuity. Video streams will be combined with preoperative MRI to track anatomical structures and their deformation. In coordination with an associated postdoctoral project, the work will integrate automatic segmentation and 3D video-to-MRI registration tools, using a dataset currently under construction that targets approximately 100 patients and combines manually annotated RGB views and corresponding MRI regions. The second strand will investigate 6D pose estimation of surgical instruments from their visible parts and known geometry. Instrument tip trajectories and velocity changes will be analysed to investigate resection cavity reconstruction, including partially occluded regions, followed by registration to MRI. These developments will support anatomical and temporal localisation of surgical actions and the generation of interactive postoperative reports, with potential applications in surgical training. Validation will first be performed on a deformable brain phantom subjected to controlled brain shift. Accuracy will be assessed against an optoelectronic motion capture system with submillimetre precision. A subsequent operating-room proof of concept is envisaged, subject to satisfactory laboratory results and completion of the regulatory preparations for the pilot study.

所需个人能力与技能： 要求具备计算机视觉与图像处理相关知识，尤其关注多模态配准、时间追踪及形变建模方面能力。在立体视觉、相机标定、3D重建、深度学习或医学影像领域具有经验者将更具优势。 该项目需要具备算法开发与实验研究方面的兴趣，包括设计可穿戴视觉系统、处理多模态数据以及进行定量方法评估。拥有视频采集或视觉系统整合经验将受到欢迎，但并非必需。 科学严谨性、独立能力、好奇心以及参与多学科团队工作的能力均至关重要。成功申请者应能够与计算机科学家和神经外科医生进行沟通，并表现出对临床应用的兴趣。在文献综述、撰写论文及开展研究汇报时，需具备良好的书面与口语科学英语能力。 要求的法语水平：高级；您能够以更复杂、更自然的方式进行表达，并涉及多种主题。 所需英语水平：高级中级：您能够有效使用语言，并能准确表达观点。