论文部分内容阅读
目的结合四维CT(four-dimension computed tomography,4D-CT)技术,应用自由变形模型(free form deformation,FFD)形变配准算法分析呼吸运动对原发性肝癌患者肝脏及其靶区形态的影响。方法选取2010-08-01-2011-08-30山东省肿瘤医院8例接受放疗的原发性肝癌患者,获得患者4D-CT的模拟定位图像,将4D-CT图像依据呼吸周期均分为CT0、CT10、CT20、CT30、CT40、CT50、CT60、CT70、CT80和CT9010套CT图像。以呼气末时相图像CT50为参考图像,其他9幅图像采用基于B样条的自由变形模型FFD算法与其配准。然后用差分位图、Hausdroff距离及互信息技术分析呼吸运动对肝脏及其肿瘤形态的影响。结果从一个完整呼吸周期10个时相的CT图像及差分位图可以看出,呼吸运动对正常肝脏及肿瘤靶区形态造成的影响,其中吸气末和呼气末的变化最大。基于B样条的多分辨率网格自由变形模型FFD形变配准算法,可有效实现不同CT图像间的形变配准。其将肝脏和肿瘤靶区配准前CT0与CT50的Hausdroff距离由0.091 4减小到0.043 2,平均减小了54.1%;互信息从0.789 0提高到0.814 0;平均提高了3.17%。结论呼吸运动造成的肝脏及其肿瘤靶区的形变是显著的,基于B样条的多分辨率网格自由变形模型FFD形变配准算法可有效实现4D-CT图像间的形变配准。
Objective To analyze the effect of respiratory motion on the morphology of liver and its target in patients with primary liver cancer by using four-dimensional computed tomography (4D-CT) technique and free-form deformation (FFD) deformation registration algorithm. Methods 2010-08-01-2011-08-30 Shandong Province Tumor Hospital 8 cases of patients with primary liver cancer undergoing radiotherapy obtained 4D-CT simulated positioning images of patients, the 4D-CT images were divided into respiratory cycle based on CT0 , CT10, CT20, CT30, CT40, CT50, CT60, CT70, CT80 and CT9010 sets of CT images. The images of end-stage CT50 were taken as the reference images, and the other nine images were registered with the FFD algorithm based on the B-spline free-form model. Difference bitmap, Hausdroff distance and mutual information technique were then used to analyze the effects of respiratory motion on the liver and its tumor morphology. Results CT images and differential bitmaps at 10 phases of a complete respiratory cycle showed that respiratory motion had the most effect on morphology of normal liver and tumor target. Based on the B-spline multi-resolution grid freeform FFD deformation registration algorithm, the deformation registration between different CT images can be effectively achieved. It reduced the Hausdroff distance of CT0 and CT50 from 0.091 4 to 0.043 2 before the registration of liver and tumor targets, with an average reduction of 54.1%. The mutual information increased from 0.789 0 to 0.814 0, with an average increase of 3.17%. Conclusions The deformation of the liver and its tumor target area caused by respiratory movement is significant. The FFD deformation registration algorithm based on B-spline multi-resolution grid free deformation model can effectively achieve the deformation registration between 4D-CT images.