放疗中基于灰度的图像配准算法性能研究

Performance Research of Gray-scale Image Registration Algorithm in Radiotherapy

目的:阐述图像配准在放疗中应用的关键问题,对基于灰度的3种配准方法的性能做深入研究,包括均方测度、归一化相关测度以及互信息测度。方法:分析各配准要素的算法原理后,基于C++加以实现,提出使用综合配准误差来评价不同配准算法的性能,并与传统目标配准误差的评价结果作对比。结果:3种测度都能对近模态的图像实施准确的配准,其中互信息测度驱动的配准在配准精度和速度上表现更为稳定,综合配准误差仅为另外两个测度的一半左右。结论:利用综合配准误差得到的评价结果更为客观,互信息测度是放疗中实施配准的较理想测度。

Objective To introduce the key issues of image registration in radiotherapy and study the performances of the three kinds of registry ways based on gray-scale registration, included mean-square measure, normalized correlation measure and mutual information measure. Methods After analysis of elements of the matching algorithm, the different registration algorithm performance are evaluated by the use of a comprehensive registration error based on C ＋＋ to be realized, which can compare with the evaluation results of traditional target registration error. Results All three kinds of measure can both match accurately in plesiotype model images, which registration accuracy and speed in the standard of mutual information measure is more stable, integrated registration error is only half of the other two measures. Conclusion The comprehensive evaluation result by using registration error is more objective, the mutual information measure is better registration measure in radiotherapy.