基于体素灰度三维多模医学图像配准中相似性测度的选取(英文)

Choice of Similarity Measure in Voxel Intensity Based 3D Multi-modal Medical Image Registration

目的在基于体素灰度医学图像配准领域 ,找出最适合于临床应用的多模医学图像配准相似性测度。方法在极端的刚体配准条件下 ,检验出互相关系数 ,互信息和相关比相似性测度为适合的相似性测度。同时进一步解释了基于互信息相似性测度的医学图像配准易于陷入局部最优 ,而基于相关比相似性测度的方法易于保证配准得到全局最优。最后 ,利用加速的多分辨率配准方案和Powell′s优化算法 ,对临床医学图像进行了基于相关比相似性测度的多模图像配准试验。结果通过临床医学专家的判断 ,利用相关比相似性测度进行多模医学图像配准 ,完全能满足临床的要求 ,进行MR/CT、MR/PET三维多模医学图像配准时效果非常理想。结论相比于其他相似性测度 ,互相关比相似性测度在基于体素灰度 ,三维多模医学图像配准领域 。

Objective To find out the most appropriate similarity measure for clinical use by comparison research on similarity measures commonly used in voxel intensity based multi modal medical image registration. Method Under extreme rigid registration condition, correlation coefficient,mutual information and correlation ratio similarity measures were tested as most suitable similarity measures. It was explained why mutual information, but not correlation ratio based medical image registration be easily trapped in the local optimization maximum. At last, using accelerated multi resolution registration scheme and powell′s optimization algorithm, experiments based on clinical medical images were implemented to evaluate the correlation ratio similarity measure for multi modal medical image registration. Result Judged by clinical expert,correlation ratio based multi modal medical image registration method worked well for clinical application, the effect of correlation ratio based, 3D MR/CT and MR/PET, multi modal medical image registration was promising for clinical use. Conclusion Compared with other similarity measures, correlation ratio is more appropriate and accurate for 3D multi modal medical image registration.