基于体素灰度3D多模医学图像配准中的相似性测度

Similarity Measures in Voxel Intensity Based3D Multi-Modal Medical Image Registration

对基于体素灰度多模医学图像配准中广泛采用的相似性测度 (SM)进行了比较研究 ,认为在配准条件极不理想的条件下 ,基于互信息 (SM)、归一化互信息 SN、相关比 (SR)的 SM是最为适用的 .分析了基于 SR 的配准法相比于 SM,易于保证配准得到全局最优变换 .利用基于 SR 的配准方法 ,对磁共振 (MR)和 CT、MR和正电子发射断层扫描 (PET)临床医学图像进行配准 。

A comparison research on similarity measures popular at voxel intensity based multi-modal medical image registration was given. Under the extreme registration condition, the mutual information, normalized mutual information, correlation ratio are considered as most suitable similarity measures. This paper also explained why the mutual information based medical image registration can easily be trapped in the local maximum in optimization process, but not for the correlation ratio. The experimental results based on clinical medical images show that the correlation ratio similarity measure based multi-modal medical image registration method work well for clinical application.