摘要
统计形状模型构建的关键在于训练集样本形状的构造和形状特征点的自动提取.针对多模态心脏图像的配准问题,提出了一种统计形状模型的构建方法.模型的构建过程主要通过图谱标签图像训练集的建立、模板标签图像形状特征点的提取和模板形状到待标记图谱形状特征点的自动传递来完成.并在此构建方法的基础上,建立左心室统计形状模型引导多模态图像配准过程.为评估构建方法的有效性,使用心脏CT和MR图像数据集进行多模态配准实验验证.结果表明,基于统计形状模型约束的方法较仅靠互信息的方法在配准精度上有明显提高.
The key of statistical shape model construction lies in the establishment of sample shape on training set and the automatic extraction of shape feature points. Aiming at the problem of multi-modal cardiac image registration, a method to build a statistical shape model is proposed in this paper. The process of model construction is mainly completed by establishing the training set of atlas label image, extraction of shape feature points in template label image and automatic transfer of feature points from template shape to the atlas shape to be marked. On the basis of this construction method, statistical shape model of the left ventricle is established to guide the multi-modal registration procedure. To assess the effectiveness of the construction method, cardiac CT and MR datasets were used to validate multi-modal registration. The experimental results show that the proposed method can improve registration accuracy compared to the method using mutual information only.
作者
陆雪松
郭翔宇
LU Xuesong;GUO Xiangyu(College of Biomedical Engineering,South⁃Central University for Nationalities,Wuhan 430074,China)
出处
《中南民族大学学报（自然科学版）》
CAS
2020年第1期67-73,共7页
Journal of South-Central University for Nationalities：Natural Science Edition
基金
湖北省自然科学基金资助项目(2016CFB489)
中央高校基本科研业务费专项资金资助项目(CZY19024)
关键词
统计形状模型
心脏图像
多模态配准
互信息
statistical shape model
cardiac image
multi-modal registration
mutual information
作者简介
陆雪松(1975-),男,副教授,博士,研究方向:医学图像分析,E-mail:xslu scuec@hotmail.com。
