中文摘要：

为了提高图像质量,提出了一种基于线性多项式拟合和主成分分析的形变向量场优化算法.对初始四维磁共振（4D MR）图像的形变向量场进行优化,然后利用新的形变场来形变参考图像,从而得到不含或者含有较少运动伪影的4D MR图像.该算法使合成的4D MR图像中肿瘤和隔膜区域的形状扭曲得到缓解,并且保持原有的运动模式.2种测量值的一致性以及对结果的定性分析表明：该算法能保存4D MR图像的运动信息,并且进一步提高图像质量,为肿瘤精确放疗提供条件.

英文摘要：

4 D MR images have been playing an important role in tumor target definition in treatment planning and tumor motion assessment during radiation therapy. To reduce motion artifacts caused by respiratory motion and further improve image quality, an effective method based on combined linear polynomial fitting and principal component analysis was proposed to remodel displacement vector fields （ DVFs） of original 4D-MRI, and the reference image was then deformed using the new DVFs to acquire 4D MR images without or reduced motion artifacts in this study. The distorted regions around the diaphragm and tumor were mitigated and maintained original motion pattern in the synthetic 4 D MR images. The consistency of two measurements and the qualitative analysis of results show that the proposed algorithm can improve the image quality of 4 D MR images on the basis of keeping original motion pattern, which provids good conditions for precise tumor radiation therapy.