Zhang Jingzhou, Li Ting, Zhang Jia
School of Automatic Control, Northwestern Polytechnical University, Xi'an 710072, China.
Sheng Wu Yi Xue Gong Cheng Xue Za Zhi. 2008 Dec;25(6):1303-6.
Medical image registration is a highlight of actual research on medical image processing. Based onsimilarity measure of Shannon entropy, a new generalized distance measurement based on Rényi entropy applied to image rigid registration is introduced and is called here generalized mutual information (GMI). It is used in three dimensional cerebral image registration experiments. The simulation results show that generalized distance measurement and Shannon entropy measurement apply to different areas; that the registration measure based o n generalized distance is a natural extension of mutual information of Shannon entropy. The results prove that generalized mutual information uses less time than simple mutual information does, and the new similarity measure manifests higher degree of consistency between the two cerebral registration images. Also, the registration results provide the clinical diagnoses with more important references. In conclusion, generalized mutual information has satisfied the demands of clinical application to a wide extent.
医学图像配准是医学图像处理实际研究中的一个亮点。基于香农熵的相似性度量，引入了一种基于雷尼熵的新广义距离度量并应用于图像刚性配准，在此将其称为广义互信息（GMI）。它被用于三维脑部图像配准实验。仿真结果表明，广义距离度量和香农熵度量适用于不同领域；基于广义距离的配准度量是香农熵互信息的自然扩展。结果证明，广义互信息比简单互信息用时更少，并且新的相似性度量在两个脑部配准图像之间表现出更高的一致性程度。此外，配准结果为临床诊断提供了更重要的参考。总之，广义互信息在很大程度上满足了临床应用的需求。
