ESTRO 2020 Abstract book

S1030 ESTRO 2020

robust a regularisation term was used to limit large variations in parameter values of successive gantry angles.

reviewed whether plan changes were made after peer review. Results 934 SBRT cases that underwent peer review were identified. The most common treatment sites were lung (518, 55%), liver (196, 21%) and spine (119, 13%). Learning moments were identified in 161 cases (17%) and plan changes were made after peer review in 28 cases (3%). Two critical errors (0.2%) were identified: an inadequate PTV margin, and an incorrect image set used for contouring. There was a statistically significantly higher rate of learning moments for lower volume SBRT sites (defined as £100 cases, including bone, lymph nodes, adrenal, kidney, and pancreas) versus higher volume SBRT sites (lung, liver, spine; 29% vs 16% respectively; p=0.001). The most common learning moments were related to overall practice/management (25%), target delineation (23%), dose/fractionation (23%), and treatment technique (17%). There was no difference between the types of learning moments nor the rate of learning moments for SBRT cases compared to non-SBRT radical cases over the same time frame (17% among 2610 cases, p=0.98). Conclusion Peer review is a critical aspect of SBRT treatment. The most common benefit appears to be significant discussion contributing to team learning, practice improvement and standardization. There are also direct benefits to patient safety and quality of care through plan revisions and critical error identification. All SBRT sites benefit from peer review, but lower volume sites may require particular attention. PO-1758 Performance studies of an MR linear accelerator using a 3D-printed ball-bearing phantom H.L. Riis 1 , B.C. Buthler 1 , U. Bernchou 1,2 , A.S. Bertelsen 1 , S.N. Agergaard 1 , V.N. Hansen 1 , F. Mahmood 2 , C. Brink 1,2 1 Odense University Hospital, Department of oncology, Odense, Denmark ; 2 University of Southern Denmark, Department of clinical research, Odense, Denmark Purpose or Objective The aim of this work was to develop a method for quality assurance of geometrical performance on MR linear accelerators (linac) at low cost, fast, and simple in use. The proposed method should enable measurement of the geometric positions of the radiation source and imager panel (EPID) as a function of gantry angle, including possible rotation uncertainties of the panel, in order to quantify the geometric performance of the linac. Material and Methods A cylindrical phantom of outer diameter of 180 mm, wall thickness 5.4 mm, length 200 mm were printed on a 3D- printer in polylactic acid (PLA) plastic, see Fig. 1, including an EPID image of the phantom. Steel ball-bearings with diameter 3.96 mm were mounted 30° apart in the wall of the cylinder forming two rings with 12 ball-bearings in each. Each ring was parallel and separated by 70 mm. To optimise the EPID image quality at the MR linac, the bridge and the couch were removed from the beam. Additional support for the phantom was used to position it at the isocentre height, i.e. to the centre of the bore. The linac was characterised in terms of the following degrees of freedom: source position (3), EPID panel position (3), EPID rotations (3) in total nine parameters. The aim was to extract these nine gantry dependent parameters from the EPID images of the phantom. The position parameters are defined in IEC 61217, thus z parallel to beam direction; x and y orthogonal to z . The 24 ball-bearings within the phantom results in additional 72 parameters to include residual uncertainty of ball positions in the phantom. Over a 360° gantry rotation 72 EPID images were acquired. A cost function, using all 72 images, was created to minimising distance between observed and expected ball positions in the EPIDs. To make the optimisation more

Results A series of 72 images acquired on an Elekta Unity MR linac was analysed. In Fig. 2a, the source position and the imager (img) panel position variations are shown as function of gantry angle. The variations are generally small all less than 1 mm. Also the rotation variation of the panel (not shown) was very limited. For the sake of comparison, also image acquisition of the phantom on a conventional Elekta Synergy linac was carried out, Fig. 2b. Largest variations are seen in the source ( x ) position. Comparison of Fig. 2a and 2b show a much better performance of the MR linac compared to a conventional linac. The variation of the conventional linac was found in agreement with findings in the literature.

Conclusion