基于快速自由形变的弹性配准算法在放疗CT中的应用

Application of fast free-form elasticregistration algorihm in radiotherapy CT

介绍了一种基于快速自由形变的弹性配准算法,并将其应用到放疗患者不同分次间的CT图像中,为进一步分析肿瘤及危及器官的变化及累加剂量提供技术支持。通过Matlab软件编写三维全自动弹性配准程序,应用此程序对鼻咽癌放疗患者不同分次时的两组三维CT图像进行仿真实验验证,并与基于传统Lucas-Kanade光流场模型的三维弹性配准相比较。基于快速自由形变的弹性配准算法配准前后对比,最小均方误(MSE)减少了19.8%,相关系数(CC)提高了7.1%;Lucas-Kanade光流场模型算法配准前后对比,MSE减少了37.8%,CC提高10.6%。两种配准方法均取得较好的配准效果,基于传统的Lucas-Kanade光流场模型的三维弹性配准的评价参数好于快速自由形变配准法算法,但快速自由形变配准法算法配准时间只是传统的Lucas-Kanade光流场模型的0.576倍。不过上述两种算法对于细微差别处,仍显得精度不够,如要应用到在线自适应放疗中仍需进一步提高形变的精度与进一步减少配准时间。

A fast free-form elastic registration algorithm was presented in this paper, and applied in radiotherapy CT image, pr- oviding technical support for further analysis of the changes of tumor and organs at risk, and evaluating cumulative dose during radiotherapy. A 3D fully automatic image elastic registration program was written by matlab software, and applied to do simulated experiment verification for two groups of 3D CT images of patients with NPC, and then compared with 3D elastic registration based on the traditional Lucas-Kauade optical flow model. The comparison between before and after the registration of fast free-form elastic registration algorithm showed that minimum mean square error （MSE） decreased by 19.8%, and that eorrelatien coefficients （CC） increased by 7.1%. For 3D elastic registration based on the traditional Lucas-Kanade optical flow model, MSE decreased by 37.8% and CC increased by 10.6%. Both these two registration methods had achieved good results. Although the evaluation parametex of 3D elastic registration based on Lueas-Kanade mode was a bit better than fast free-form algorithm, the registration time of the latter just was 0.576 times of that of the former. The above two algorithms are not precise enough for subtle, so these algorithm~ still need to further improve accuracy and reduce registration time in the appfication to the online adaptive radiotherapy.