基于自回归模型功率谱估计预测放射治疗中靶区的呼吸运动

Prediction of respiratory motion based on power spectrum estimation of autoregressive model during radiotherapy

目的:基于自回归模型功率谱估计提出一种预测放射治疗中靶区呼吸运动的方法,为解决实时跟踪放疗中系统延迟问题提供参考。方法:通过X射线连续透视影像获取肺癌患者靶区呼吸运动信号,利用自回归模型功率谱估计的Burg方法预测靶区呼吸运动,比较预测误差E_(predicted)和系统延迟误差E_(shifted),分析预测算法的有效性。结果:用Burg方法预测10例肿瘤患者呼吸运动信号,E_(predicted)=(0.6±0.3)mm,E_(shifted)=(2.5±1.3)mm,E_(predicted)<E_(shifted)(P<0.01)。结论:用Burg方法预测靶区呼吸运动是有效的,可以作为临床运用的一种选择。

Objective To compensate the time lag in real-time tracking radiotherapy by proposing a method for predicting the respiratory motion of target volume during the radiotherapy based on the power spectrum estimation of autoregressive （AR） model. Methods Tumor respiratory motion signals of patients with lung cancer were obtained through consecutive X-ray fluoroscopy images. The respiratory motion of target volume was predicted by Burg method of AR model. The validity of Burg method was evaluated by comparing predicted error （Epredicted） and the error of time lag （Eshitied）. Results Ten tumor respiratory motion signals were predicted by Burg method. According to the results, Epredicted = （0.6±0.3） mm, Esbifted = （2.5±1.3） ram, Eptedictedd 〈 Eshifted （P〈0.01）. Conclusion Burg method which can be an option for clinical application is effective in tumor respiratory motion prediction.