ESTRO 2020 Abstract book

S1047 ESTRO 2020

Results Registration deviation from the ground truth physical shift in different LR and SI positions was shown in Fig. 1. The mean and SD of registration deviations were well below 1mm at all phantom positions. No trend of registration deviation increase was observed for different phantom positions. The maximum registration deviations at different locations to the ground truth physical shift and calculated systematic and random errors were presented in table 1. The group means of registration deviations was 0.1mm and -0.05mm in LR and SI directions; the systematic ± random errors in LR and SI direction were 0.06 ± 0.53 and 0.08 ± 0.49 mm, respectively. No significant difference was observed of registration deviation in LR ( p = 0.0946) or SI ( p = 0.0967) direction using Wilcoxon signed rank-sum test.

we implemented an inhouse developed software script to calculate two values: one that is representative of the integral dose delivered to the patient, and another one which is a measure for the complexity of the plan. The integral dose measure is compared to the one in the original plan. Based on the complexity score (ranging from 0-1, 0 for highly complex, 1 for open fields) we decide whether to perform a measurement of treatment plan with an IMRT verification phantom (Delta 4, Scandidos). To ensure that the electron densities are correctly assigned to the new MRI, we compare also the equivalent pathlength of the original and the adapted plan. Results For the first 70 patients, and 268 adaptive fractions, the independent MC calculations showed an average 2%/2mm gamma passing rate of 97.8% (Range 91.2% - 99.8%). For 93% of the plans the independent point dose calculation agreed within 10% of the one from the original plan (average 1.1%, range -16.3% - 22.8%). Most values above 10% were connected to lung treatments, where the scatter is incorrectly taken into account in the point dose calculation. In 95% of online adaptions, the difference in the values representing the integral dose comparing the adapted plan with the original one was below 10% (average difference 1.4%, range -18% - 22%). Complexity scores ranged between 0.11 and 0.62 (average 0.33). For complexities below 0.4 we performed a measurement after the patient’s treatment. All measurements of the original plan passed the gamma-analysis with a 3%/3mm criterion with on average 99.9% (range 97.8% - 100%). The adapted plans had equally good passing rates (average 99.8%, range 97.2% - 100%). The equivalent pathlength differed more than 2cm in 21 fractions, where the changes could be traced back to changes in anatomy. Conclusion We developed and implemented a comprehensive and vendor independent QA chain for the online adaptive planning process on the MRIdian MR Linac which ensures a safe treatment of the patient. PO-1785 Image registration evaluation of online Monaco on Unity MR-Linac - A phantom study Y. Zhou 1 , J. Yuan 1 , Y.W. Ho 1 , W.W. Lam 1 , B. Yang 1 , L.C. Ho 1 , K.Y. Cheung 1 , S.K. Yu 1 1 Hong Kong Sanatorium & Hospital, Medical Physics and Research Department, Hong Kong, Hong Kong SAR China Purpose or Objective To evaluate the rigid image registration performance of online-MONACO on Unity MR-Linac through a phantom study. Material and Methods A CIRS 3D abdominal phantom (model 057A) was laterally aligned on the MR-Linac treatment table at index 24, serving as the reference position. The phantom was then manually shifted along LR and SI with an increment of 0.5cm within the most probable online shift range in our hospital (±25mm around the isocenter), yielding total positions of 121 (LR shift = 5× i ; SI shift = 5× j in mm; where i, j = -5 to 5, resulting in a total positions of 11×11=121). Phantom shift along LR was manually measured by a ruler that was taped to the front of the phantom, and shift along SI was conducted by console table position control. MR images were acquired at each position using the default MR-LINAC pelvis protocol of a 3D T2 turbo-spin-echo sequence (TR/TE = 1535/278ms, reconstructed voxel size = 1.5 x 1.5 x 2 mm, scan time = 117s) and were rigidly registered to the images acquired at the reference position using online Monaco (version 5.4). Mean, SD, systematic error (Σ) and random error (σ) of image registration deviation from the ground truth physical shift were calculated. Wilcoxon signed rank sum test was used to assess the registration deviation at different phantom shift locations along LR and SI directions with a significant level of p = 0.05.

Fig. 1 Average registration deviations in LR and SI at all shift locations. Error bar represents the standard deviation.