医学图象匹配的遗传算法方法  

GENETIC OPTIMIZATION METHODS FOR MEDICAL IMAGE REGISTRATION

作　　者：段会龙[1] 贾春光[1] 谭鸥[1] 吕维雪[1] 

机构地区：[1]浙江大学生物医学工程研究所,杭州310027

出　　处：《中国生物医学工程学报》2001年第4期364-370,379,共8页Chinese Journal of Biomedical Engineering

摘　　要：本文用变形轮廓线的方法实现医学图象匹配过程中的对应标记点问题 ,并用遗传算法进行最佳匹配结果的搜索。遗传算法用变形轮廓的能量函数作为适应度函数 ,并通过动态映射方法实现搜索空间尺度的自适应变化。实验结果表明 ,本文的方法能有效地用于建立医学图谱和图象之间的一一对应关系 ,是对以往人工交互标注方法的重要改进。We solved the problem of homogenous landmarks in the procedure of medical image registration via deformable contour model, and genetic algorithm (GA) was used to search the optimal results. The GA used the energy function of the deformable contour as its fitness function, and an adaptive search space scaling technique was presented which was operated by dynamic mapping. The experimental results showed that the proposed method could efficiently establish the correspondence between medical atlas and medical images, which was an important improvement to the previous interactive methods.

关 键 词：医学图象匹配 图象分割 遗传算法 

分 类 号：R445[医药卫生—影像医学与核医学] TP391.41[医药卫生—诊断学]