基于Viscous Fluid模型的快速CT-CBCT图像变形配准算法研究  

The Study of CT to CBCT Deformable Image Registration Based on Viscous Fluid Model

作　　者：王琳婧[1] 张书旭[1] 林生趣[1] 余辉[1] 彭莹莹[1] 周露[1] 

机构地区：[1]广州医学院附属肿瘤医院放疗中心,广东广州510095

出　　处：《中国医疗设备》2013年第6期21-23,33,共4页China Medical Devices

基　　金：国家自然基金项目(81170078);广东省科技计划项目(2011B031800111);广州市科技计划项目(2011J4300131)

摘　　要：CT和CBCT图像之间的变形配准是自适应放射治疗中的一项关键技术。本研究利用互信息量作为相似性测度,基于Viscous Fluid模型,在图形处理器GPU框架下用CUDA编程语言实现了CT-CBCT快速图像变形配准。本研究选用6例临床头颈部肿瘤患者的CT和CBCT图像对本文算法进行了验证和分析,并与经典demons算法进行了对比。实验结果表明,本文算法能快速、精确地完成CT和CBCT图像的变形配准,配准时间为秒级,能较好地满足临床对于快速变形图像配准的要求。CT to Cone-beam CT （CBCT） deformable image registration （DIR） is one of the key techniques in adaptive radiotherapy. Based on the Viscous Fluid model, it uses the mutual information as a similarity measure and achieves the CT to CBCT deformable image registration with CUDA language under the GPU framework. In this study, our algorithm was validated and analyzed by the CT and CBCT images of six clinical head-and-neck cancer patients, and was compared with the classic demons algorithm. Results showed that our algorithm can complete the deformation registration of CT and CBCT images quickly and accurately, which can meet the clinical requirements for fast deformable image registration.

关 键 词：CT图像 CBCT图像 图像变形配准 图形处理器 

分 类 号：R319[医药卫生—基础医学] TP391.41[自动化与计算机技术—计算机应用技术]