显式重定向的扩散张量图像配准

REGISTRATION OF DIFFUSION TENSOR IMAGES BASED ON EXPLICIT REORIENTATION

扩散张量图像配准算法是近年图像配准研究的热点与难点之一。针对配准中容易出现的局部极值和张量重定向问题,以欧氏距离为相似性测度,将张量重定向显式融入目标函数,采用模拟退火算法与Powell算法法相结合的混合优化策略,对临床使用的扩散张量图像DTI(Diffusion Tensor Images)进行配准实验。实验结果表明,该算法稳定性良好,在对扩散张量图像进行配准时,能有效保持扩散张量主特征方向与纤维走向的一致性,同时成功解决了局部极值的困扰,是一种实用的扩散张量图像配准方法。

Registration of diffusion tensor images is one of the hotspots and nodi in image registration studies in recent years. In light of the problem that during the registration the local extrema are easily to happen and the problem of tensor reorientation, in this paper the Euclidean distance was taken as the similarity metric, and the tension reorientation was explicitly fused into the target function, by applying a novel hybrid optimization strategy which conjoined the simulated annealing algorithm with Powell algorithm, the registration experiment on diffusion tensor images （DTI） for clinical use were carried out. Experimental results shown that this algorithm has a stable performance, it can effectively ensure the principal orientations of diffusion tensor remaining consistent with the fibre direction whiles resolve the perplexing of local extrema successfully when registering DTL It is a practical registration approach for DTI.