免疫ICP算法在三维医学图像刚性配准中的应用研究

发布时间：2018-05-29 12:05

本文选题：医学图像配准 + 医学图像变换　；参考：《河北工业大学》2015年硕士论文

【摘要】：近年来,随着医学影像技术的不断进步和发展,各种模态的医学图像应运而生。临床诊断应用中,为了使多种模态的图像信息可以得到综合利用,就需要把不同模态或者不同时刻、不同成像条件的医学图像结合起来。但是,不同模态的图像在空间位置上往往不能保证一致,因此首先要进行图像配准,运用多模态医学图像配准技术来解决这一问题。医学图像配准是指通过寻找某种空间变换,使两幅图像的特征点达到空间位置和解剖结构上的完全一致,要求配准的结果能使两幅图像上所有的解剖点或至少是具有诊断意义以及在手术区域内的点都达到匹配。迭代最近点(Iterative Closest Point,ICP)算法是一种基于特征点集的图像刚性配准算法。在一系列刚性配准算法中,ICP算法能使两幅图像之间的距离快速达到很近的位置,但是这一最近距离值收敛到一定的程度便不再变化。因此,为了提高配准的精度,本文结合了免疫算法来对ICP算法的配准结果精度进行进一步的提高。本文分别运用ICP算法、免疫算法和免疫ICP算法对二维和三维医学图像进行了不同变换程度下的配准实验,并且对三种算法的实验结果进行了对比分析。ICP算法收敛速度快,稳定性也较好,但仅使用ICP算法获得的配准精度到达一定值之后便不再提高。免疫算法的收敛速度比较慢,不适合单独使用。本文采用的免疫ICP算法结合了两种算法的优点,获得了较好的配准效果。实验结果表明,免疫ICP算法在两套二维图像数据源实验中的配准精度比ICP算法提高了11.96%和11.86%,在两套三维图像数据源实验中的配准精度比ICP算法提高了12.51%和10.75%。
[Abstract]:In recent years, with the continuous progress and development of medical imaging technology, various modes of medical images emerge as the times require. In clinical diagnosis, in order to make the image information of multiple modes be used synthetically, it is necessary to combine the medical images with different modes, different times and different imaging conditions. However, the spatial location of images with different modes is often not consistent. Therefore, the first step is to carry out image registration, and to solve this problem by using multi-modal medical image registration technology. Medical image registration means that the feature points of two images can reach the same spatial position and anatomical structure by looking for some kind of spatial transformation. The registration results can match all the anatomical points or at least the points in the surgical area of the two images. Iterative nearest Closest algorithm is a rigid image registration algorithm based on feature set. In a series of rigid registration algorithms, ICP algorithm can make the distance between two images reach a very close position quickly, but the nearest distance value converges to a certain extent and does not change. Therefore, in order to improve the accuracy of registration, this paper combines the immune algorithm to further improve the accuracy of the registration results of the ICP algorithm. In this paper, ICP algorithm, immune algorithm and immune ICP algorithm are used to carry out registration experiments of 2D and 3D medical images in different degrees, and the experimental results of the three algorithms are compared and analyzed. The convergence speed of the three algorithms is fast. The stability is also good, but the registration accuracy obtained by using ICP algorithm can not be improved when it reaches a certain value. The convergence speed of immune algorithm is slow, and it is not suitable for use alone. The immune ICP algorithm proposed in this paper combines the advantages of the two algorithms and achieves a good registration effect. The experimental results show that the registration accuracy of immune ICP algorithm in two sets of two-dimensional image data source experiments is 11.96% and 11.86% higher than that of ICP algorithm, and 12.51% and 10.75% higher than that of ICP algorithm in two sets of 3D image data source experiments.
【学位授予单位】：河北工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.41

【参考文献】