Nonrigid 3D Medical Image Registration and Fusion Based on Deformable Models
Nonrigid 3D Medical Image Registration and Fusion Based on Deformable Models
复制标题
基于可变形模型的非刚性3D医学图像配准与融合
DOI：
10.1155/2013/902470
复制
发表时间：
2013
影响因子：
--
通讯作者：
L. Lüdemann
中科院分区：
文献类型：
--
作者：
Peng Liu;Benjamin Eberhardt;C. Wybranski;J. Ricke;L. Lüdemann
对于医学图像的配准，刚性方法往往不能提供足够的自由度，而可靠的弹性方法仅适用于临床上的特殊应用。弹性模型的自由度数必须减少，以便在临床环境中使用，以存档可靠的结果。提出了一种新的基于几何的非刚体三维医学图像配准与融合方法。该方法以基于三维曲面的可变形模型为指导。在我们的双重方法中，首先将来自其中一幅图像的可变形网格应用于待配准对象的边界。之后，从曲面网格点的位移推断出在活动曲面所描述的感兴趣区域(ROI)内进行配准和融合所需的非刚性体积变形矢量场。使用准刚性器官(肾脏)和弹性器官(肝脏)的临床图像对该方法进行了验证。参考图像和模型之间的图像强度差的标准差的减小被用作性能的度量。放置在肝脏血管分叉处的标志物被用作评估弹性肝脏注册结果的金标准。将我们的配准方法与基于互信息的仿射配准方法应用于准刚性肾脏进行了比较。对于刚性配准，新方法的质量提高了15.11%，置信度高达99%。然而，当应用于准弹性肝脏时，该方法的平均标志性位移量为4.32 mm。相比之下，摘取肝脏的仿射配准产生的错位明显更小(P=0.000001)，为3.26 mm。总之，我们的验证表明，新方法适用于内部变形不是关键的情况，但在也必须考虑内部位移的情况下具有局限性。
For coregistration of medical images, rigid methods often fail to provide enough freedom, while reliable elastic methods are available clinically for special applications only. The number of degrees of freedom of elastic models must be reduced for use in the clinical setting to archive a reliable result. We propose a novel geometry-based method of nonrigid 3D medical image registration and fusion. The proposed method uses a 3D surface-based deformable model as guidance. In our twofold approach, the deformable mesh from one of the images is first applied to the boundary of the object to be registered. Thereafter, the non-rigid volume deformation vector field needed for registration and fusion inside of the region of interest (ROI) described by the active surface is inferred from the displacement of the surface mesh points. The method was validated using clinical images of a quasirigid organ (kidney) and of an elastic organ (liver). The reduction in standard deviation of the image intensity difference between reference image and model was used as a measure of performance. Landmarks placed at vessel bifurcations in the liver were used as a gold standard for evaluating registration results for the elastic liver. Our registration method was compared with affine registration using mutual information applied to the quasi-rigid kidney. The new method achieved 15.11% better quality with a high confidence level of 99% for rigid registration. However, when applied to the quasi-elastic liver, the method has an averaged landmark dislocation of 4.32 mm. In contrast, affine registration of extracted livers yields a significantly (P = 0.000001) smaller dislocation of 3.26 mm. In conclusion, our validation shows that the novel approach is applicable in cases where internal deformation is not crucial, but it has limitations in cases where internal displacement must also be taken into account.