ESTRO 38 Abstract book

S986 ESTRO 38

Twenty-two prostate cancer patients were randomly selected, all of them with seminal vesicles and a simultaneous integrated boost to the prostate. For each patient, two identical VMAT plans were created in Eclipse v13.7 using 2 arcs and 6 MV photons. One of the plans was optimised using PO v13.7 and the other using PRO v13.7 with the same optimisation goals. The dose distribution was calculated with Acuros XB in terms of dose-to-water and normalised to the median dose to the prostate. The dosimetric comparison of both plans was done for the PTVs by using Conformity Index (CI), Homogeneity Index (HI) and S-index, D98%, D50%, and D2%. Doses to OAR were also compared for the rectum (V44, V61, V65), bladder (V61, V65), and femoral heads (D2%, V50). Plan complexity was evaluated with a software developed in MATLAB by our National Society that returns several complexity indices and parameters: MUs, MLC information (speed and acceleration, gap and its variation speed), tongue-and- groove effect (T&G), interdigitation, Modulation Index Total, Modulation Complexity Score, Beam Irregularity, leaf travel, and dose rate and gantry speed variations. Optimisation times were also compared. The statistical analysis of the paired differences was performed using Wilcoxon tests. Results The dosimetric differences are shown in Figure 1 as violin plots (median, quartiles, and density for each algorithm).

Optimisation times were lower for PO (42%) than for PRO. Plan complexity was handled differently. PO plans had more MU and higher dose rate variations, total modulation, and MLC speed. Nevertheless, the MLC shape was simpler for PO: lower Beam Irregularity, higher and more stable (lower variation speed) mean gap between opposed leaves, and lower T&G. The interdigitation was higher for PO due to the pairs of leaves partially included in the limits in Y direction: PO moves its touching point outside the jaws whereas PRO puts it in the middle. There were no gantry speed variations in any case. Conclusion PO and PRO optimisation algorithms implemented in Eclipse were analysed for prostate VMAT treatments in terms of dosimetric quality and plan complexity. PO optimisation times were much lower than PRO’s. There were no clinically relevant differences in the dose distributions, but plan complexity was handled differently, with PO moving it from MLC shape to MLC speed and dose rate variations. Overall, PO is preferable over PRO for prostate VMAT treatments. EP-1819 Robustness to shifts in patient position with either knowledge based or multi-criteria optimisation P. Houston 1 , D. Keiller 1 1 Beatson West of Scotland Cancer Centre, Radiotherapy, Glasgow, United Kingdom Purpose or Objective To perform a comparison of the robustness of two treatment planning optimisation methods to shifts in patient position for VMAT Pancreas Planning. The optimisation methods investigated were a knowledge based planning solution, RapidPlan (RP v15.5), and a multi-criteria optimisation solution (MCO v15.5). Material and Methods Knowledge based planning was performed with RP. Multi- criteria optimisation was performed with MCO. All plans were two arc axial VMAT and were normalised to a PTV mean of 100% dose. RP generates estimated DVHs and planning objectives based on the library of plans it is built from. MCO is a planning trade off tool that calculates multiple plans for each planning objective; this allows the planner to explore the Pareto surface of the ideal dose distribution to

All these differences were not clinically relevant (below 1%). Optimisation time and plan complexity differences are shown in Figure 2.