利用数学形态学工具配准人脑MR-PET图像  

HUMAN BRAIN MR-PET IMAGE REGISTRATION USING MATHEMATICS MORPHOLOGICAL TOOLS

作　　者：杨虎[1] 马斌荣[1] 任海萍[2] 沈晋慧[1] 

机构地区：[1]首都医科大学BME系,北京100054 [2]中国医学科学院肿瘤医院核医学科,北京100005

出　　处：《中国生物医学工程学报》2002年第5期430-436,455,共8页Chinese Journal of Biomedical Engineering

摘　　要：数学形态学是以形态为基础对图像进行分析的数学工具。我们利用简单的形态学操作从PET和MR图像中提取对应的解剖结构皮质层 ,以归一化互信息为相似性测度 ,进行医学PET/MR图像的几何对准。评估结果证明 ,该方法的配准精度能达到亚像素精度 ,并能极大地节省运行时间 ,是一种稳健性强、精度高。Mathematic morphological tools are used for image analysis based on morphology. We applied simple morphological operation to extract corresponding anatomic structurecortex, and chose normalized mutual information as similarity measure for the geometric alignment of primate PET and MR images. Preliminary evaluation results showed that sub voxel accuracy could be achieved with this approach and the registration progress could also be speedup exceedingly. The new algorithm we presented is a very robust and automated multi modality registration method with high accuracy.

关 键 词：数学形态学 配准 互信息 人脑 MR-PET图像 

分 类 号：R318.04[医药卫生—生物医学工程]