基于增强CT图像的肝脏血管三维分割方法研究  

3D Hepatic Vascular Segmentation on Contrast-Enhanced CT Images

作　　者：张睿 吴水才[1] 周著黄[1] 王月 吴薇薇 ZHANG Rui;WU Shuicai;ZHOU Zhuhuang;WANG Yue;WU Weiwei(College of Life Science and Bioengineering,Beijing University of Technology, Beijing 100124, China;Faculty of Information Technology, Beijing University of Technology, Beijing 100124, China)

机构地区：[1]北京工业大学生命科学与生物工程学院,北京100124 [2]北京工业大学信息学部,北京100124

出　　处：《中国医疗设备》2017年第11期48-54,共7页China Medical Devices

基　　金：国家自然科学基金项目(71661167001);北京市属高校青年拔尖人才培养计划项目(CIT&TCD201404053);中国博士后科学基金项目(2017M620566)

摘　　要：目的本研究旨在探寻一类性能优异的血管增强算法,并结合阈值水平集分割算法进行肝脏血管系统的三维自动分割。方法首先对原始三维增强CT数据进行S型非线性灰度映射;随后对不同的血管增强算法进行对比分析;最后使用阈值水平集分割算法分割出肝血管系统。选用3Dircadb公开数据集中的20例腹部增强CT数据定量评估了两类经典的血管增强算法,包括血管特征提取算法及扩散滤波算法。结果血管特征提取算法运行效率平均优于扩散滤波算法。血管特征提取算法结果的对比度平均高于扩散滤波算法2 d B以上,导致扩散滤波算法后续的计算复杂度高,准确性降低。阈值水平集分割算法的结果与区域生长算法、形态检测水平集算法和测地线活动轮廓水平集算法相比,准确性达77%以上,高于其余分割算法。结论血管特征提取算法与扩散滤波算法相比,更适合依赖灰度值的血管分割。阈值水平集算法能缓解单纯依赖阈值或依赖血管边界的血管欠分割问题,结合血管增强算法后能更准确的分割出肝脏血管。Objective To investigate a set of state-of-the-art vessel enhancement algorithms and propose an automatic3D hepatic vascular segmentation strategy by combining the vessel enhancement and the threshold level set segmentation algorithms.Methods Firstly,the sigmoid filter was conducted to the original3D contrast-enhanced CT images.Then,different vessel enhancement algorithms were analyzed and compared on the filtered images.Finally,the liver vessel was segmented by combining the threshold level set segmentation method with the optimal vessel enhancement algorithm.Two groups of vessel enhancement algorithms were quantitatively assessed:vesselness filters and diffusion filters.Experimental data involved20cases of abdominal contrast-enhanced CT in the3Dircadb public data sets.Results Experimental results showed that the vesselness filters were more effective than the diffusion filters.The contrast-to-noise ratio of the vesselness filters was higher than that of the diffusion filters by2dB.Thus,the diffusion filters yielded higher computational complexity and lower accuracy.The segmentation accuracy of the proposed method was over77%,which was higher than that of traditional segmentation algorithms including region growing,shape detection level set and geodesic active contours.Conclusion Compared with the diffusion filters,the vesselness filters are more suitable for grey level based on3D hepatic vascular segmentation.The threshold level set method could alleviate the under-segmentation problem when using only one threshold or vessel boundaries,and could segment hepatic vessels more accurately with the combination of vessel enhancement algorithms.

关 键 词：肝血管分割 血管增强 扩散滤波 血管特征提取 阈值水平集分割 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]