基于多对比度磁共振颈动脉图像的配准算法研究

发布时间：2018-01-09 10:02

本文关键词：基于多对比度磁共振颈动脉图像的配准算法研究　出处：《中国生物医学工程学报》2017年02期 　论文类型：期刊论文

【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。
[Abstract]:In recent years, multi-contrast magnetic resonance imaging has become a powerful tool for plaque analysis. However, its performance is affected by the mismatch of blood vessels in multi-sequence images. In order to achieve accurate blood vessel registration of multi-sequence images. Based on the segmentation of the lumen, a two-step registration strategy from coarse to fine is proposed. Firstly, the rigid registration of the center line of multi-contrast image is realized by iterative nearest point. In the second step, in order to find the matching points of different blood vessel boundary accurately, the shape context descriptor is used to filter the boundary points. Deterministic annealing technique is applied to global optimization and a new 3D multi-contrast magnetic resonance angiography sequence is used to quantitatively evaluate the effectiveness of the proposed algorithm. After registration, the boundary coincidence degree of different sequences is more than 95% and the average surface distance is 0.12 mm, which can effectively improve the registration accuracy and lay a foundation for the subsequent analysis of plaque composition.
【作者单位】：第四军医大学生物医学工程系;清华大学生物医学影像研究中心;
【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)
【分类号】：R445.2;TP391.41
【正文快照】： 引言大量临床研究表明,颈动脉粥样硬化斑块成分及易损性与脑卒中等心脑血管疾病的发生密切相关[1],对颈动脉斑块及其成分的诊断分析成为目前影像学技术研究的热点。其中,核磁共振成像具有无创、无辐射、软组织分辨率高等优势,特别是,多对比度MRI成像既能够显示管腔狭窄,又能显

【相似文献】