基于差分自回归移动平均模型的医用直线加速器剂量偏移预测研究

Study on Prediction of Medical Linear Accelerator Dose Offset Based on Auto Regressive Integrated Moving Average Model

目的结合时间序列挖掘中对未来值的预测方法,实现医用直线加速器质控数据偏移的预测。方法选取1个调整周期内前19条医用直线加速器剂量监测数据作为时间序列的观测组,随后的5条医用直线加速器剂量监测数据作为参照组,建立前19条监测数据的时间序列,对时间序列进行稳定性分析,确定构建时间序列模型——差分自回归移动平均(ARIMA)模型,并利用模型对其他调整周期内的监测数据进行预测。结果利用时间序列进行医用直线加速器剂量偏移预测的表现良好,预测值与实测值的对比误差为-1.28%~0.80%,偏移量总体趋势相同。结论该研究提出的医用直线加速器剂量偏移预测方法对加速器剂量参数的及时调整起到了参考和提示作用。

Objective To predict the offset of quality control data of medical linear accelerator in combination with the future value predicting method used in time series data mining.Methods The first 19 records and the subsequent 5 records of dose monitoring data of medical linear accelerator within an adjustment period were considered as the observation group and the reference group of time series,respectively.Then,a time series set up for the first 19 records of monitoring data was analyzed over its stability.On this basis,the time series model—auto regressive integrated moving average(ARIMA)model was established and adopted for prediction of monitoring data within other adjustment periods.Results The errors between the predicted and the measured values ranged from-1.28%to 0.80%,indicating that the dose offset of medical linear accelerator can be well predicted with the time series,and the general tendency of offset is consistent.Conclusion The proposed method of predicting dose offset of medical linear accelerator can provide a reference and suggestion for timely adjusting dose parameters for the accelerator.