基于改进单纯形的互信息配准方法研究

Study of Mutual Information Registration Based on Modified Simplex Optimization Method

由于成像机理不同和人体组织结构的高度复杂性,单模态的医学图像不能提供医生所需要的足够信息。多模态医学图像的配准和融合有着十分重要的意义。图像配准是图像融合首先要解决的问题。配准的目的是使两幅图像的位置在空间上达到一致。对MRI和PET两幅图像进行配准,先采用主轴法对两幅图像进行粗略的配准,利用两幅图的互信息作为相似性量度,采用改进单纯形法进行全局搜索,实现最佳配准。结果表明,采用由粗到细的配准策略和改进单纯形的优化搜索算法,配准精度高,计算速度快,可完成不同分辨率下多模态图像的精确匹配。

Because of a different imaging mechanism and highly complexity of body tissues and structures, single -modality medical image can not provide enough information for clinical doctors. This has very important signifi- cance for multimodal medical image registration and fuse. Image registration is the first and key part of problem to be solved in the integrations. When the spatial position of two medical images is same, the registration could be a- chieved. A mutual information registration based on modified simplex optimization method is presented in this paper to improve the speed of medical image registration. For two MRI and PET images, the principal axis method is adopted to achieve the rough registration, and the modified simplex algorithm is employed to implement global search using the mutual information as similarity measure. Results indicate that the proposed registration method prevents the opti- mizing process from falling into local extremum and improves the convergence speed while keeping the precision. The accurate registration of muhimodal image with different re^olntions is achieved.