基于CBCT的三维/四维可视化介入影像导航关键问题的研究

Recently many image-guided Interventional techniques have been well developed and increasely accepted in medical practice, 3 dimensional image guidance is obviously prefered to 2D guidance during these interventional procedures. Computer and software assisted guidance navigation technique greatly improved the accuracy, speed and safety of intervention treatment based on 3D data set. Cone beam CT （CBCT）constructed by 3D image acquisition via a X-ray panel detector has been increasing available in the world.CBCT function was applied to increase the accuracy and speed of interventional procedure in some aspects. but many disadvantages were found during this CBCT image guidance, such as, low density resulation of CBCT, difficulty of many target tissue segmentation, poor visulization of target area on fused 3D image, much influence by respiratory movement,etc. Based on the past one-year NSFC study project by us, the purpose of this study is to furtherly explore and stuy the key issues on computer assisted visualized fluoroscopy navigation guidance system via X-ray flat panel detector, such as, to study the feasibity of image fusion of CBCT with pre-precedure diagnostic CE-MSCT to increase ability of segmentation of interest tissue structures; to explore the fusion method to maximally revise the organ movement and figuration change caused by different test; to explore the respratory phase chase technique so that to revise the error which caused by respiratory movement during the fusion of realtime 2D X-ray fluoroscopy and 3D volume data. This study would obviously increase the accuray and speed of traditional interventional procedures, and axpand its usage to area of prior CT guidance interventions,such as tumor biopsy and ablations; further more,increase the target segmention and reduce the X-ray exposure, compared with traditional plain CT guidance.

计算机辅助下三维影像导航技术在现代影像引导的介入诊疗领域逐渐发挥优势。大平板X线探测器在国内外逐渐普及，其CBCT三维成像功能在介入引导中起到一定作用，但存在诸多问题影响其性能的进一步提升，如CBCT密度分辨率低，靶区分割困难，呼吸影响影像配准，可视化性能差等。本课题将在我们过去一年期研究课题的基础上进一步研究CBCT精确可视化引导的关键科学问题，诸如如关联融合术前增强螺旋CT提高CBCT的靶区分割能力、融合算法研究最大程度纠正MSCT和CBCT两种检查因体位和呼吸所致器官位置改变和形变；探索呼吸时相追踪技术问题纠正呼吸所致肝脏形变和位移引起的配准误差等等，以实现平板探测器的的三维或四维可视化介入影像导航。该研究将明显提高平板探测器引导传统介入诊疗的准确性和安全性，并拓展其引导领域到CT引导介入穿刺领域，且在一定程度上优于CT平扫的介入引导，比如增加靶区的识别能力和减少术中X线曝光剂量等。

在2013年度一年期国自然研究的基础上，进一步针对CBCT三维和四维可视化影像导航技术关键科学基础问题和方法的研究。构建一个多模态信息融合的手术模型，利用3D增强CT影像中肝脏，靶区以及肝血管的分割算法研究，从而在弱对比影像上实现三维的可视化显示。提出了一种增强软组织弱边界的对比项，改进了现有的图像割算法中对弱边界的分割能力，实现肝脏的准确分割，平均体积交叠比误差仅为5.3%。设计了一种基于灰度和形状校正的自适应似然估测算法实现肝脏肿瘤的准确和稳健的半自动分割，Dice均值达到了84%。设计了一种基于3D UNet的深度网络实现血管特征自动提取与分类的算法，Dice均值达到了75%。探索肿瘤碘油沉积三维CBCT和二维透视显示差异，研究发现Lip-CBCT (Az=0.75)优于二维透视图像 (Az=0.54)。采用跟踪类膈肌呼吸特征的Amsterdam Shroud方法与感兴趣区域内密度分析的方法实现了病人特异性呼吸信号的提取方法，改善呼吸运动过程中组织配准错误的问题。利用增强现实技术开展三维和四维可视化透视导航系统研究，搭建了基于Qt的肝脏介入手术规划和治疗系统的软件平台，辅助医生实现CT引导的肝介入手术方案的规划；探索了利用投影仪的可视化引导，通过模型实验验证。探索晚期肝门胆管癌的介入治疗价值，提高患者中位OS到20.5个月。