Huang Xishi, Zaheer Sameer, Abdalbari Anwar, Looi Thomas, Ren Jing, Drake James
Department of Medical Imaging, University of Toronto and CIGITI, Hospital for Sick Children, Toronto, Canada.
Annu Int Conf IEEE Eng Med Biol Soc. 2012;2012:2347-50. doi: 10.1109/EMBC.2012.6346434.
Fast extraction of blood vessels of abdominal organs is still a challenging task especially in intra-procedural treatments due to large tissue deformation. In this study, we propose a novel joint vessel extraction and registration framework. This vessel extraction technique is under the guidance of prior knowledge patient specific models. The proposed technique automatically provides correspondence between extracted vessels and pre-procedural vessels, which is important for image guidance such as labeled vessels from pre-procedural models, improves the quality of disease diagnosis using multiple images and follow-up, and provides important information for nonrigid image registration. Another key component in our framework is to dynamically update mapped pre-procedural models by rapidly registering the patient model to the current image based on strain energy, point marks and 3D extracted vessels currently available. We have demonstrated the effectiveness of our technique in extraction of vessels from liver MR images. Validation shows a extraction error of 3.99 mm. This technique has the potential to significantly improve the quality of intra-procedural image guidance, diagnosis of disease and treatment planning.
快速提取腹部器官血管仍然是一项具有挑战性的任务，特别是在术中治疗中，因为组织变形较大。在本研究中，我们提出了一种新颖的联合血管提取和配准框架。这种血管提取技术是在特定患者模型的先验知识指导下进行的。所提出的技术自动提供提取血管与术前血管之间的对应关系，这对于图像引导（如术前模型中的标记血管）很重要，提高了使用多幅图像进行疾病诊断和随访的质量，并为非刚性图像配准提供重要信息。我们框架中的另一个关键组件是通过基于应变能、点标记和当前可用的3D提取血管将患者模型快速配准到当前图像，动态更新映射的术前模型。我们已经证明了我们的技术在从肝脏磁共振图像中提取血管方面的有效性。验证显示提取误差为3.99毫米。该技术有可能显著提高术中图像引导、疾病诊断和治疗计划的质量。
