小波分解结合自适应神经模糊推理系统的呼吸预测研究

Respiration prediction by wavelet decomposition and adaptive neuro fuzzy inference system

目的:研究一种方法精确预测胸腹部肿瘤放射治疗中的非规则呼吸运动。方法:提出基于小波分解和自适应神经模糊推理系统的呼吸运动预测方法(WANFIS),利用小波分解将呼吸信号分成基线、低频和高频三部分,并分别采用线性拟合、自适应神经模糊推理系统(ANFIS)、简单移动平均进行预测,然后综合三部分预测值作为呼吸运动预测结果。基于30例临床数据回顾性分析,将WANFIS算法与神经网络(NN)、CyberKnife放射外科系统的Synchrony呼吸同步追踪系统、ANFIS这三种典型预测算法进行对照比较。结果:本文提出的WANFIS算法的归一化均方根误差(nRMSE)平均值为0.09,小于NN的0.17、Synchrony的0.11以及ANFIS的0.11。结论:WANFIS能更好地预测非规则呼吸信号,更有效地补偿放疗系统时间延迟。

Objective： To develop a method to precisely predict irregular respiratory motion in radiotherapy for thoracic and abdominal tumors. Methods： A prediction algorithm based on wavelet decomposition and adaptive neuro fuzzy inference system （WANFIS） was proposed. The respiratory signal was first decomposed into baseline, low frequency and high frequency components, which were then predicted respectively using linear fitting, adaptive neuro fuzzy inference system （ANFIS） and simple moving average. The three parts of predicted values were finally combined to form the respiration prediction result. The WANFIS was compared with three typical prediction algorithms including neural network （NN）, Synchrony Respiratory Tracking System of the CyberKnife Robotic Radiosurgery System and ANFIS by retrospective analysis of clinical data of 30 cases. Results： The normalized root mean square error （nRMSE） of WANFIS averaged 0.09, which was less than 0.17 for NN, 0.11 for Synchrony and 0.11 for ANFIS. Conclusion： The WANFIS algorithm proposed in this paper may more precisely predict irregular respiratory motion and thus more effectively compensate the time delay of the system.