基于B样条自由形变模型的医学图像非刚性配准算法研究

Non-rigid Registration Algorithm Based on B-spline Free Medical Images of the Deformation Model

在放疗临床应用中，病人器官组织的蠕动变形经常使得计划的靶区与实际放疗不符。非刚性配准技术作为解决这一问题的有效手段，已成为目前医学图像处理领域的研究热点和难点。本文针对乳腺癌治疗中MR图像肿瘤病变区的定位，研究了基于B样条的自由形变模型进行图像非刚性配准，并采用有限记忆L—BFGS优化算法对代价函数进行优化，降低了计算量和存储量，缩短了优化时间。实验结果表明，该算法不仅效率高，而且配准效果好。

In the clinical application of radiation therapy, the patient＇s deformation of organ organization often make planned target section not agree with actual radiation therapy. Non-rigid registration technology as the effective solution to the problem, has become the hot and difficult point of medical image processing field. This paper aim at location of MR image lesions area in breast cancer treatment, use free form deformation model based on the B-spline to implement image non-rigid registration. This method utilizes limited memory L-BFGS to optimize the cost function,which reduces computational complexity and memory capacitance, and shorten the optimization time. The experimental results show that this method is not only of high efficiency, but also of good effect in registration.