A Comparison of Adaptive Boosting Algorithms for the Respiratory Signal Prediction

Dynamic tracking of tumor with radiation beam in radiation therapy requires prediction of real-time target location ahead of beam delivery as the treatment beam with beam or gating tracking brings in time latency. We proposed an adaptive boosting (Adaboost) method based on the multi-layer neural network (MLP-NN) to predict the respiratory signal accuracy in our previous study. Recently, several variants of Adaboost methods showed great potential for the regression prediction problem. Hence, we investigated the prediction performance of four popular adaptive boosting method based on the MLP-NN for the respiratory prediction problem in this study. The root-mean-square-error (RMSE), correlation coefficient (CC) and maximum error (ME) between predicted and real respiratory signals, obtained from the Real-time Position Management, were used to evaluated the prediction performance. The Adaboost.RT method and the Adaboost method used in our previous study get the best RMSE and CC while the Adaboost. BCC obtained the best ME. The experiment results demonstrated that appropriate Adaboost method based on MLP-NN could predict the respiratory signal accuracy.