Respiratory Motion Prediction Based on Maximum Posterior Probability

For the radiotherapy, the tumor inside thorax or abdomen keep varying with respiration motion. Current technologies, e.g., respiratory gating and beam tracking, face great challenges in predicting the respiratory tumor motion. Whereas respiratory motion is changeful, traditional prediction model such as Linear Model, Kalman Filter, and so on, can not imitate the motion accurately. In this article, the probabilistic algorithm, combined with the state inference, is proposed in order to predict the respiration signal during treatment. The respiratory objects of eleven patients were employed in our work to validate the proposed method. The experimental results were satisfying in comparing with traditional methods, e.g., the method successfully dealed with various local variations in respiratory objects, and predicted the respiration with lower error and higher correctness rate of state inference, so much as the signals with different time latency.