基于多对比度磁共振颈动脉图像的配准算法研究  

Multi-Contrast Magnetic Resonance Imaging Registration of Carotid Arteries

作　　者：吴玉霞[1] 徐肖攀[1] 张曦[1] 刘洋[1] 张国鹏[1] 陈慧军[2] 卢虹冰[1] 李宝娟[1] 

机构地区：[1]第四军医大学生物医学工程系,西安710032 [2]清华大学生物医学影像研究中心,北京100084

出　　处：《中国生物医学工程学报》2017年第2期129-135,共7页Chinese Journal of Biomedical Engineering

基　　金：国家自然科学基金青年基金(81301199);国家自然科学基金重点项目(81230035);陕西省社会发展科技攻关项目(2016SF302)

摘　　要：基于在血管壁增强显示方面的强大潜力,近年来多对比度核磁共振成像已成为斑块分析研究的有力工具,但其效能受到多序列图像血管不匹配的影响。为实现多序列图像的准确血管配准,在管腔分割的基础上提出一种由粗到精的两步配准策略:先采用迭代最近点实现多对比度图像中心线的刚性配准,再采用薄板样条实现基于血管边界点的非刚性配准。在第二步配准中,为准确寻找不同序列血管边界的匹配点,创新使用形状上下文描述子对边界点进行筛选,并应用确定性退火技术进行全局优化。采用新型三维多对比度磁共振血管成像序列,对提出算法的有效性进行定量评价。结果表明,配准后不同序列的边界重合度均达到95%以上,平均表面距离0.12 mm,可有效提高配准精度,为后续斑块成分的分析奠定基础。Recently multi-contrast magnetic resonance imaging （MRI） has become a powerful tool for plaque analyzing due to its strong potential in enhanced demonstration of carotid wall, hut its performance is hampered by the misalignment of difference imaging sequences. To achieve accurate vessel registration of multi-sequence images, a two-step coarse-to-fine registration strategy was proposed in this study. First, the iterative closest point was applied to realize the rigid registration of the centerlines of the multi-contrast images, and then the thin plate spline was used to realize the non-rigid registration based on the boundaries of carotid artery. In the second step, to find the corresponding points of the boundaries in different sequences, a shape context descriptor was innovatively introduced to screen the boundary points. In addition, the deterministic annealing technique was employed to find a globally optimized solution. The effectiveness of the proposed algorithm was quantitatively evaluated by novel three-dimensional （3D） multi-contrast vessel wall MRI sequences. The results indicated that after registration, the overlap of two boundaries from different sequences was more than 95% , and their mean surface distance was 0. 12 mm, which improved the accuracy of registration effectively and laid the foundation for further component analysis.

关 键 词：颈动脉粥样硬化 易损斑块 磁共振图像 配准 

分 类 号：R318[医药卫生—生物医学工程]