2D／3D图像配准中的相似性测度和优化算法

摘要

在手术引导治疗中，2D/3D图像配准能辅助医生准确定位病人病灶，而准确的配准涉及相似性测度和优化算法等众多方面。为了研究相似性测度和优化算法对2D/3D图像刚性配准的影响，本文结合6种相似性测度和4种优化方法在配准“金标准”数据上进行了2D/3D图像配准实验，并从配准成功率、平均迭代次数和平均配准时间三个方面对配准结果进行了对比研究。实验结果表明，以模式强度为相似性测度，用Powell方法进行优化搜索是最佳配准组合。并且，在不改变相似性测度条件下，Powell方法是所用优化方法中配准效果最好的优化方法。%In surgical guide treatment,2D/3D medical image registration can provide the precise position of patient for surgeon.Accurate registration involves many aspects,such as similarity measurements and optimization methods.In or-der to investigate the influence of similarity measurements and optimization methods on 2D/3D image registration,a comparison of six similarity measurements in combination with four optimization methods is performed using the public and available porcine skull phantom datasets from Medical University Vienna.Comparison is performed for the regis-tration results based on success rate,the number of iterations and execution time.The results show that the most accu-racy registration is obtained by pattern intensity combined with Powell.Furthermore,the best 2D/3D registration re-sults are obtained by Powell search strategy with fixed similarity measurement.