中文摘要：

目的研究图像配准算法的基本行为，分析影响图像配准的各种因素。方法先把待配准图像做一已知变换，然后基于自主开发的图像配准算法研究图像配准技术中感兴趣区域选择、优化目标函数时的初始位置选择、空间变换参数的空间耦合等对图像配准算法的影响。结果感兴趣区域位置和区域大小选择对图像配准的结果影响非常明显，如果选择不当很容易导致配准失败；初始变换参数（优化的初始位置）选择恰当，能提高图像配准的准确度；参数空间的耦合会增加图像配准算法对感兴趣区域选择的敏感度。结论图像引导放疗平台临床应用中非常有必要总结与肿瘤部位相关感必趣区域的勾画策略（依赖具体图像配准软件平台），可以通过实现初始变换参数的自动优化选择来提高算法的抗干扰能力，并开发出针对不同图像特征（不同解剖部位）的具体配准算法。

英文摘要：

Objective Study the behaviors of image registration algorithms, and analyze the elements which influence the performance of image registrations. Methods Pre-known corresponding coordinates were appointed for reference image and moving image, and then the influence of region of interest （ROI） selection, transformation function initial parameters and coupled parameter spaces on registration re- sults were studied with a software platform developed in home. Results Region of interest selection had a manifest influence on registration performance. An improperly chosen ROI resulted in a bad registration. Transformation function initial parameters selection based on pre-known information could improve the accu- racy of image registration. Coupled parameter spaces would enhance the dependence of image registration algorithm on ROI selection. Conclusions It is necessary for clinic IGRT to obtain a ROI selection strategy （depending on specific commercial software） correlated to tumor sites. Three suggestions for image registra- tion technique developers are automatic selection of the initial parameters of transformation function based on pre-known information, developing specific image registration algorithm for specific image feature, and assembling real - time image registration algorithms according to tumor sites selected by software user.