面向癫痫深部电极植入手术的多模态MRA脑血管网络构建

Deep electrode implantation is a minimally invasive and accurate surgery for epilepsy localization. However, intracranial hemorrhage would be likely to occur in the process of operation if blood vessels are punctured. Currently, Magnetic Resonance Imaging Angiography (MRA) is one of the most commonly used cerebrovascular imaging methods in clinical practice. In this research, we plan to construct the whole cerebrovascular network by segmentation and integration of noncontrast-enhanced multimodal MRA dataset, which can be applied for some patients who are prone to nephrogenic systemic fibrosis in the process of contrast agent injection. The constructed cerebrovascular structures would provide security for the epilepsy deep electrode implantation surgery. The research contents are as follows: (1) Due to grayscale of tiny blood vessels is low in PCA and they are hardly to be detected. The vascular grayscale and shape features are fused using Dempster–Shafer evidence theory to segment the cerebrovascular structures from PCA image. (2) Since the background of TOF image is complex and the gray scale distribution of blood vessels is large, an automatic vascular seed point detection method is proposed, and the vascular characteristics image is constructed based on fuzzy inference to segment the cerebrovascular structures from TOF image. (3) In order to verify whether the construed cerebrovascular structures of noncontrast-enhanced multimodal MRA can be applied to the deep electrode implantation surgery, imaging and anatomical verification will be carried out. The research results can not only provide the basis for the surgical planning of minimally invasive neurosurgery, but also provide theoretical and method support for the development of multimodal MRA imaging equipment.

癫痫深部电极置入手术是一种微创的、精准的癫痫灶定位方式，然而手术过程中易发生血管被刺破而颅内出血的情况。MRA是目前临床上较为常用的脑血管成像方式，本课题拟通过研究多模态MRA影像中血管的分割与融合，实现全脑血管结构网络的构建，为癫痫深部电极置入手术提供安全保证。研究内容如下：（1）针对PCA影像中小血管亮度较暗不容易被检测的问题，基于D-S证据理论实现血管灰度特征和形状特征的融合，分割出PCA影像中脑血管的结构；（2）针对TOF影像中背景复杂、血管灰度分布范围较大的特点，提出血管种子点自动检测方法，基于模糊推理获得的血管特征融合图像，分割出TOF影像中脑血管的结构；（3）为了验证构建的脑血管结构能否应用于手术，本课题将对多模态MRA影像中构建的脑血管结构进行影像学和解剖学的验证。课题研究成果不仅可以为微创神经外科的手术规划提供依据，还可以为研制多模态MRA影像设备提供理论与方法支持。

多模态磁共振血管造影（Magnetic Resonance Imaging Angiography, MRA）的三维血管网络结构的高精度模型重建，能够提高医生对脑血管信息的准确认知，辅助医生对癫痫深部电极置入手术的路径规划，有效提高手术成功率、减少脑出血等并发症。本项目基于癫痫深部电极置入手术的血管网络构建所涉及的理论问题和关键技术进行研究，主要研究内容包括：通过研究相位对比法血管造影（phase-contrast Angiography, PCA）影像和时间飞跃法（time-of-flight, TOF）影像的结构特征，建立了基于数学模型、机器学习以及深度卷积等多方法交叉及结合的方法体系，实现了针对不同应用场景及需求的多种方法的血管特征提取及三维重建；最后，通过研究多模态MRA中分割出的脑血管结构与对比增强法血管造影（contrast-enhanced, CEA）中分割的血管和开颅后大脑皮层血管的对比，实现算法的影像学验证和解剖学验证。.项目的主要目标是围绕癫痫深部电极置入手术中的脑血管成像技术中的关键和难点问题的解决提供新理论、新方法和新技术，在脑血管的特征图像构造、自动提取、精确分割等关键技术的研究上取得突破性进展。项目所涉及的脑血管的分割三维神经网络结构和其他关键性技术是其他器官和病灶组织分割（心脏组织血管、肝血管组织、病理性肺组织、新冠肺炎感染区域等）的共性问题和核心技术，具有重要的理论研究意义；所构建的增强现实系统在人机交互的基础上，成功投入到抗击新冠疫情行动中，在当下这种突发性、爆发式的传染性疾病上体现了该研究的价值。