Abstract ▼ AI-Helper

In radiotherapy, especially in heavy ion therapy, the accuracy of the treatment beams delivered to lung tumor is very important because patient breathing causes a motion of a tumor resulting in decrease of radiation delivered to the tumor. Furthermore, one of the most problems is the systematic late...

In radiotherapy, especially in heavy ion therapy, the accuracy of the treatment beams delivered to lung tumor is very important because patient breathing causes a motion of a tumor resulting in decrease of radiation delivered to the tumor. Furthermore, one of the most problems is the systematic latencies between the acquisition of the target position and the mechanical response of the system to follow the target position. To compensate the aforementioned latencies while tracing the tumor motion, an accurate algorithm to predict these latencies is therefore required. In this study we address the problem of predicting the position of lung tumor motion during heavy ion therapy treatment in which the couch (robotic bed system) is adjusted to align the patient to the beam ― a technique used for real-time tracking in radiotherapy. The motion of lung tumor will be simulated in real-time by using single axis actuator ― Respiratory Motion Phantom (RMP) while the markers of tracking system (NDI Polaris Spectra) are set up on the RMP and recorded with 20Hz sampling rate. The marker position was smoothed by the à trous wavelet decomposition that was provided as input to predictor to predict the position for 800 ms in advance. Two prediction algorithms were implemented: Normalized Least Mean Square (NLMS) prediction which is one of the best conventional prediction algorithms, and Predicted Error-based Normalized Least Mean Square (PE-NLMS) prediction which is proposed to improve the tracking accuracy. Both algorithms were evaluated by using the first prototype robotic bed in one-dimensional motion. The results show that the PE-NLMS is outperformed with respect to Root Mean Square Error (RMSE) by NLMS and does not require too much parameter adjustment.

주제어