Error Of Multimodal Image Registration In MR-Guided Adaptive Radiotherapy And Its Effect On Dose

Magnetic resonance imaging guided radiotherapy(MRgRT)is a new development direction and hot spot in the field of radiotherapy.Compared with cone beam computed tomography(CBCT)guided technology,MR imaging has the advantages of clear imaging and high contrast resolution of soft tissue,and can be used to evaluate the effect of radiotherapy through a variety of different functional imaging sequences.As MRgRT is widely used,it becomes more important to explore the influence of different factors on MR-CT registration error in the process of MRgRT treatment.The Elekta Unity system is a magnetic resonance guided linac system which combines a 1.5T MRI system and a nominal energy 7MV beam.There is always a high magnetic field in the Unity treatment room,so it is impossible to install the laser lamp equipped with the traditional accelerator for precise setup.Therefore,in order to improve the accuracy of error measurement,the EPID system is used to replace the laser system for auxiliary setup.Firstly,a phantom that can be fixed on the radiotherapy bed board was used in the study to verify the accuracy of the EPID-assisted set-up method.The specific methods are as follows: CT simulation positioning imaging was performed on the phantom,and then the relative position of the phantom was kept unchanged and placed on the Unity bed for EPID scanning.It was observed that the position of the phantom in the image did not change,which proved that the positioning method was accurate.Next,a QA model to verify the accuracy of Unity registration is proposed according to this setup method,and the specific methods are as follows: Firstly,the CT image of the phantom was scanned,and the phantom was placed on the Unity bed board for precise positioning assisted by EPID,and then the MR Image was scanned.After automatic registration of the MR And CT images,the registration shift distance was recorded.In addition,another phantom with a surface marked with a setup error marker line and which can be fitted with a dosimeter probe was also used for the end-to-end testing of Unity through the proposed registration QA model.According to the test results,MR-CT registration during the Unity treatment was within the clinical acceptance range and the dose difference of the target volume was not more than 5%.In order to explore the influence of different factors on multi-modal registration errors in MRgRT,this study used a simulated thoracic phantom with an edited motion mode to explore the influence of three factors,including setup position,motion mode and ROI selection,on Unity registration errors and dose through the registration detection QA model described above.Five kinds of setup positions,eight kinds of motion modes and two kinds of motion modes were used respectively,and 80 groups of error data under different conditions were measured by combining each other.After each registration,the reference plan was optimized according to the registration results,and the GTV dose was measured by the installed dose meter.The minimum and maximum values of ΔX are-2.30 mm and 2.10 mm,respectively,and the size range of the error is 4.40 mm.The minimum and maximum values of ΔY are-5.60 mm and 8.00 mm,respectively,and the error range is 13.60 mm.The minimum and maximum values of ΔZ are-2.30 mm and-0.90 mm,respectively,and the size range of the registration error is 1.40 mm.ΔY has the largest standard deviation of 2.32 mm and the largest fluctuation range in all data,showing the worst registration results,while ΔZ has the best stability with a standard deviation of 0.32 mm,which is smaller thanΔX and ΔY.The error obtained by registration with spine as ROI is better than that obtained by registration with GTV as ROI,so it is not reliable to only focus on the registration results of tumors in the registration stage in the actual treatment.Different motion states have different effects on the error,but the registration results of moving targets are worse than those of stationary targets,so the motion amplitude of the irradiated area should be reduced as much as possible in the actual treatment.And it is not recommended that the margin of GTV-PTV be expanded less than 3mm.According to the dose results obtained by irradiation of GTV under different conditions,it can be seen that even though MR May not be fully included in all motion trajectories of GTV,the dose results obtained by registration with spine as ROI are still excellent.