基于互信息和模拟退火算法的多模式医学图像配准

Multimodality medical image registration based on mutual information and simulated annealing

目的 实现 CT- MR多模式医学图像配准 .方法 提出了一种基于互信息和模拟退火算法的医学图像配准方法 .对多模式医学图像 ,运用信息论的基本概念“互信息(MI)”作为匹配标准 ,使用改进的模拟退火算法作为优化准则 ,通过求解两幅图像的最大互信息量搜索出最佳匹配参数 ,对待配准图像进行变换 ,从而达到配准的目的 .结果 我们把此方法应用于 CT和 MR图像的三维刚性配准 ,取得了较好的配准效果 .结论 该方法匹配精度和鲁棒性比较高 ,其中模拟退火算法克服了其他搜索算法容易陷入局部极大值的弊病 。

AIM To perform CT MR registration. METHODS A multimodality medical image registration method based on mutual information and simulated annealing was presented. When dealing with multimodality medical images, we, by using the basic conception of 'MI' in information theory as the matching criterion and simulated annealing as optimizing guide line, searched the best matching parameter through maximization of MI. Through which transformation, the registraion could be achieved. RESULTS We applied this method to the 3D rigidity registration of CT and MR and achieved good result. CONCLUSION The method has high registration precision and robust quality, while simulated annealing algorithm overcomes the disadvantages that other searching methods easily get into local maxima. So this method is an efficient registration method.