Abstract Book

S1063

ESTRO 37

geometry a RP-plan was calculated, no intervention was allowed during the RP-optimisation process. Results Both EP-plans and RP-plans provided a good dose coverage of the target volume. Little difference was observed in the homogeneity index (between 0.07 and 0.10 for both optimisation techniques) and in maximum as well as in the near-to-maximum dose. The RP-plans allowed a significantly better sparing of the organ at risks, particularly of bladder, bowel and genitals, compared to the EP-plans. The reduction of the mean dose was of 3 Gy for both bladder and bowel and of 4 Gy for the genitals. Conclusion The knowledge-based optimisation process for the planning of anal canal cancer resulted in a significantly improved plan quality compared to the experienced planner approach. This may also lead to large efficiency gains in therapy planning. EP-1956 Prostrating ourselves to the model: A comparison of anatomy specific prostate DVH estimation models M. Anderson 1 , S. Currie 1 , L. Girvan 1 1 Beatson West of Scotland Cancer Centre, Radiotherapy Physics, Glasgow, United Kingdom Purpose or Objective DVH estimation models have already been found to lead to improvement in plan quality for prostate EBRT [1]. It has previously been found that model outliers have minimal effect on plan quality [2], however, the effect from the range of plans included has not yet been explored. A key OAR in prostate EBRT is the bladder, an organ capable of significant size variation. The aim of this study was to investigate the effect of using different DVH estimation models based on bladder size on plan quality. Material and Methods Three RapidPlan™ (Varian, Palo Alto, USA) DVH estimation models were created, a non-specific bladder size model (A) (n = 149), followed by a small bladder model from which study-sets with bladders over 150 cm 3 had been removed (B) (n = 73), and a large bladder model from which study-sets with bladders smaller than 200 cm 3 had been removed (C) (n = 61). One EBRT plan was produced for each model using the Eclipse TPS (Varian, Palo Alto, USA) for twenty patient study-sets unknown to the models, which had been sorted into a large bladder (>200 cm 3 ) group and a small bladder (<150 cm 3 ) group. The plans were compared for target coverage, OAR dose, plan modulation, and calculation time. Statistical significance was tested for using a t-test with a p value ≤ 0.05 indicating significance. Results No significant difference was found for any recorded metric except for total calculation time, with the bladder size specific models taking significantly longer than the opposing bladder size model (p = 0.002 for both cases). Primary target coverage was investigated in relation to mean bladder and rectum doses with no significant relationship found. While not significant, a general trend was observed in that using Model C on the small bladder test patients tended to produce plans with lower target coverage, and vice versa. Conclusion These results indicate that there is no clinical benefit in using a bladder size specific DVH estimation model for the planning of prostate EBRT treatments. The results also suggest the possibility of using more time to produce a plan of the same quality when the right model is used, or a lesser quality plan if the wrong model is used. As the bladder size specific models were based on the original, non bladder size specific model, the results also suggest

that plan quality for prostate plans does not increase with increasing size of DVH estimation model. [1] Fogliata A, et al. On the pre-clinical validation of a commercial model-based optimisation engine: Application to volumetric modulated arc therapy for patients with lung or prostate cancer. Radiother Oncol 2014; 113: 385-391 [2] Delaney A, et al. Effect of dosimetric outliers on the performance of a commercial knowledge-based planning solution. Int J Radiat Oncol Biol Phys 2016; 94 (3): 469- 477 EP-1957 Proton grid therapy: LET and variable RBE- weighted dose distributions for interlaced beams J. Odén 1,2 , T. Henry 1 1 Stockholm University, Medical Radiation Physics, Stockholm, Sweden 2 RaySearch Laboratories, Department of Research, Stockholm, Sweden Purpose or Objective Grid therapy uses small and spatially fractionated radiation beams to improve the outcome of radiotherapy treatments. It relies on the biological observations that using such geometries, with unirradiated parts in between small radiation beams, have the possibility to improve the radiation tolerance of healthy tissues. Our research group has explored the possibility to clinically use interlaced proton beam grids to achieve very heterogeneous dose in the normal tissue, while keeping a high homogeneity in the target dose (Henry et al. 2016). So far, this has been explored assuming a constant RBE of 1.1. The aim of this study was to compare distributions of the LET d and the RBE-weighted dose of proton grid plans with conventional IMPT plans, assuming a constant RBE=1.1 and several variable RBE models. Material and Methods Conventional IMPT and proton grid plans were generated for four prostate cases. The PTV was defined as an isotropic 5 mm expansion of the prostate, resulting in PTV volumes between 50-215 cm 3 . For both modalities, two lateral fields were used to deliver a homogenous dose of 78 Gy (RBE) in 39 fractions to the PTV, assuming RBE=1.1. The grid plans were optimized to achieve similar dose coverage as the IMPT plans. The clinical goals for the PTV was D 98% ≥74.1 Gy (RBE) and D 2% ≤83.5 Gy (RBE), whereas the normal tissue goals were adopted from QUANTEC. In addition to RBE=1.1, the three variable RBE models (Carabe et al. 2012, Wedenberg et al. 2013 & McNamara et al. 2015) were used for plan evaluation. An α/β of 1.5 Gy was assumed for the PTV, while 3 Gy was used for the normal tissues. The LET d distribution was Monte Carlo calculated for all plans. Results All IMPT and grid plans met the clinical goals assuming RBE=1.1. The conformity and homogeneity of the PTV was slightly better for the IMPT plans compared to the grid plans (Figure 1a & b). The LET d distribution of the grid plans followed the interlaced pattern of the beams, giving rise to localized high LET regions at the PTV border. This contrasted with a more spread-out high LET region around the PTV of the IMPT plans (Figure 1c & d). Yet, the different LET d maps did not produce significantly distinctive RBE-weighted dose distributions when considering the variable RBE models, as seen in Figure 2 assuming the Wedenberg RBE model for a representative case. The PTV coverage was sufficient for both techniques with similar rectum DVHs (Figure 2c). The mean dose to the femoral heads was kept similar between the two delivering techniques, although the dose pattern within the volume naturally differed (Figure 1 & 2). This resulted in the different shapes of the femoral heads DVHs in Figure 2c.