基于空间加权互信息的非刚性医学图像配准

摘要

As the similarity metric criterion,mutual information is widely used in medical image registration field,in particular,in rigidregistration research,its higher accuracy and good robustness in registration effect have been demonstrated.However,due to the complexity ofnonlinear deformation,the registration methods based on mutual information must be thoroughly studied when used in non-rigid registrationarea.The method we proposed in this article uses B-spline based free-form deformation model to simulate nonlinear deformation of anatomicalstructure in medical image,and meanwhile according to this deformation model,and taking the influence of spatial information on registrationeffect into consideration,we improve the mutual information method in the way of spatial weight,and apply LBFGS method,which has goodeffect in large scale parameters optimisation,to optimising the registration parameters,we implement the program through programming,andvalidate and analyse the effect by experiments.Experimental results show that the proposed scheme obviously outperforms the traditionalmutual information method in registration accuracy.%互信息作为相似性测度标准在医学图像配准领域应用广泛，尤其是在刚性配准研究中，能实现较高配准精度和非常稳健的配准效果。然而由于非线性形变的复杂性，在非刚性配准领域，对基于互信息的配准方法进行深入研究势在必行。提出的方案采用基于B 样条的自由变换模型模拟医学图像中解剖结构的非线性形变，同时根据变换模型，考虑空间信息对配准效果的影响，以空间加权的方式对互信息方法进行改进。使用对大规模参数优化效果较好的LBFGS方法对配准参数进行优化，编程实现程序，并通过实验对效果进行验证和分析。实验结果显示，该方案配准精度明显优于传统的互信息方法。