中文摘要：

放射治疗是胸腹部肿瘤的常用治疗手段,但由于病人在治疗过程中的呼吸使胸腹部肿瘤产生运动位移降低了放射治疗的质量。人们希望在治疗的过程中对目标进行运动估计,以减少呼吸运动在治疗过程中的影响。BP人工神经网络由于其良好的非线性逼近特性被用于呼吸运动估计,可以较好地反映呼吸运动的发展趋势;但BP神经网络本质是梯度下降法,容易陷入局部最优。利用遗传算法（GA）和BP网络相结合的方法,遗传算法具有良好的全局搜索能力,弥补BP神经网络易于陷入局部最优的缺点。通过对9例呼吸运动信号进行对比实验,结果表明用GA-BP神经网络的预测精度高于单纯使用BP神经网络的预测精度。

英文摘要：

Radiation therapy is the conventional treatment for thoracic and abdominal tumors.Since the patients＇ respiratory movement leads to tumors dislocation,decreased therapeutic effect is observed.It is anticipated to estimate the tumor displacement extent to minimize respiratory motion effects.Owing to good approximation properties of nonlinear estimation for respiratory movement of BP（back propagation） neural network,it can reflect the variation trend of the respiratory motion.BP neural network,as its nature of gradient descent method,is easy to fall into local optimum.In this paper,we describe the combination of genetic algorithm（GA） and BP neural network,using the global search capability of GA to make up BP neural network vulnerable to the shortcomings of local optimization.Experiments on 9 cases of respiratory motion signal revealed that GA-BP neural network possesses a higher prediction accuracy than that of BP.