摘要
目的:利用单方向X射线成像定位算法来实时定位肿瘤靶区可以提高放射治疗精度,应用简便的同时能够降低对病人曝光的成像剂量。但是由于缺少沿射线束方向上的运动信息,如何有效地利用之前的运动信息来保证定位精度是需要解决的一个问题。方法:针对前列腺肿瘤,选择4种典型的实时定位算法,并使用十位患者的Calypso磁场定位数据进行模拟定位来比较它们的定位效果。这四种定位算法分别是α分布图法、两种基于高斯概率密度分布的算法和贝叶斯概率密度分布法。结果:α分布图法的rmse(均方根误差)小于5 mm,但是最大误差能达到50 mm;高斯概率分布法1的rmse小于2.6 mm,最大误差小于6 mm;高斯概率分布法2的定位结果中多数病人的rmse小于1.5 mm,最大误差小于7 mm,而有部分病人的rmse大于4.5 mm,最大误差大于30 mm;贝叶斯概率密度分布法的rmse小于2.5 mm,最大误差小于8.8 mm。结论:高斯概率密度分布法1和贝叶斯概率密度分布法相对优于其它两种;尤其是贝叶斯概率密度分布法能够最好地适用于各种类型的前列腺肿瘤运动的实时定位。
Objective： Real-time tumor location via monoscopic X-ray imaging can improve radiotherapy delivery accuracy and reduce the additional dose exposing to patients without adding extra equipment. The key problem of such localization algorithms is lack of motion information along the direction of imaging x-ray. Methods： For prostate tumor, four representative localization algorithms are selected for comparison in simulative treatment process using the Calypso monitoring data of 10 patients. The algorithms are or-map algorithm, two algorithms based on 3D Gaussian probability density, and Bayesian algorithm. Results： The rmse （root mean square error） of ct-map algorithm are less than 5 mm, but its max error can reach 50 ram. The rmse and max error of Gaussian algorithm No. 1 are less than 2.6 mm and 6 mm respectively. For most patients, the rmse and max error of Gaussian algorithm No. 2 are less than 1.5 mm and 7 mm respectively; but that of some patients are greater than 4.5 mm and 30 mm. And the rmse and max error of Bayesian algorithm are less than 2.5 mm and 8.8 mm respectively. Conclusions：The Gaussian algorithm No. 1 and the Bayesian algorithm are relatively better than the other two, especially Bayesian algorithm, which is more suitable for all kinds of prostate tumor motion.
出处
《中国医学物理学杂志》
CSCD
2013年第6期4491-4496,共6页
Chinese Journal of Medical Physics
