基于后验概率的呼吸信号预测

Prediction of Respiration Signal Based on Posterior Probability

放疗过程中,采用图像引导、呼吸门控或实时跟踪技术对受呼吸影响较大的胸腹部位肿瘤目标进行治疗时,需要对呼吸条件下目标的运动进行估计。呼吸运动具有不确定性,利用传统数学模型描述其变化规律时,无法有效处理该问题。本研究提出后验概率算法进行呼吸运动估计,并利用呼吸状态判别技术有效控制跟踪过程,以解决呼吸的非线性逼近和基线漂移等问题。实验通过对11例患者的呼吸运动进行预测,证实了所提出方法的有效性;在应对信号变化和延时等方面,后验概率估计与传统算法的比较,也取得了令人满意的效果。

In radiotherapy, it is necessary to utilize image-guided, gated treatment or real-time tracking to deal with the tumors locating in the lung or abdomen which are impacted by respiration motion. Therefore the motion of the tumor object has to be estimated during respiration. Linear prediction model is not valid to imitate the tumor motion because of the uncertainty of respiration motion. In this paper, we proposed a MAP-based method to predict respiration motion. The method can effectively deal with the practical problems of non-linear estimation and base line shifts; as well as the state-inferring technique of respiration is used to control the tracking process. In the experiment, the cases of respiration motions from 11 patients were used to validate our method. In dealing with challenging signal variation and time-delaying, the MAP-based method was compared with LP method, the obtained results were satisfying.