T. Fujii1, N. Miyamoto2, S. Takao2, C. Ling Fung1, and S. Fujitaka1; 1Hitachi, Ltd., Hitachi, Japan, 2Faculty of Engineering, Hokkaido University, Sapporo, Japan
Purpose/Objective(s): Patient anatomical changes during treatment in particle therapy can cause imprecise delivery of the irradiation dose, resulting in an increased dose to normal tissue and a reduced one to tumor tissue. This issue must be addressed in particle therapy, especially for pancreatic tumors. Additional margins are usually given to the treatment plan to account for these changes. We propose a new tracking technique named “real-time volume gating (RT-VG),” which enables gating not only using the surrogate marker position to represent the target position but also the changes in surrounding tissue along the treatment beam path—changes that affect the beam range in particle therapy. To investigate the feasibility of the RT-VG, the basic performance was tested with an anthropomorphic digital phantom. Materials/
Methods: The RT-VG technique uses a patient’s motion-model constructed from the correlation between the deformable vector field (DVF) derived from deformable image registration of 4DCT and the positions of one or more surrogate markers. The model is used to estimate the DVF to generate a reconstructed volume from the marker position during the treatment. The model parameters are derived using a partial least squares regression. The water equivalent thickness (WET) along the beam path on the reconstructed volume is calculated and used for the detection of beam range changes. In this study, we checked the estimated WET accuracy using the 4DCT test dataset (512×512×140 voxels, ten phases, SI/AP respiration motion: 20 mm/10 mm) generated by the extended NURBS-Based Cardiac-Torso (XCAT) phantom. A simulated tumor was inserted into the head of the pancreas as a target. A surrogate marker that moves with the surrounding tissue was also inserted near the target. The treatment beam angle was set to 210 degrees. The size of the volume reconstructed by the model for calculating WET changes along the beam path was 62×62×205 voxels (734 cc). The calculation time from the detection of the surrogate markers on simulated DR images to the gating judgement based on the WET change, including volume reconstruction, was also evaluated using prototype software. Results: The mean absolute error of WET estimation was 0.8 mm for all phases. The mean time of processing from the marker detection to the gating judgement, including volume reconstruction and WET calculation, was 68.4 msec per frame. Conclusion: In this phantom study, the RT-VG technique enabled checking deviation in the beam range during treatment with reasonable accuracy and computation time. We will further consider the VG criteria on the basis of the dose deviation caused by WET changes and develop an interlock function for volume gating. This method will be evaluated with actual patient data in treatment workflow using cone beam CT to correct the model for each treatment day.
