ESTRO 38 Abstract book

S1003 ESTRO 38

recalculated with a validated Monte Carlo(MC) code with a statistical uncertainty of 1%. We evaluated the effect of setup and range uncertainties on the planned dose distribution with an in-house software simulating 8 setup errors and 2 range errors (16 calculations per plan in total), to estimate near worst-case scenarios for target volume and organs at risks (OAR). We evaluated the differences between MC and PB distributions and the plan robustness using the same dosimetric indices used to optimize the plan (e.g. D1 and D99 for the target volume, D1 for serially responding OARs, etc.). Results For the intracranial lesions, the differences between PB and MC was small (mostly <1%) and setup&range error were causing the largest differences between nominal and actual(perturbed) dose. For the remaining treatment sites, CTV coverage in MC plans was always lower than in PBS plans, in some cases with difference in CTV D99 by more than 10%(Table1). The differences were the largest when a preabsorber was needed to irradiate the shallowest part of the target. Concerning the OAR, the differences between MC and PB algorithm were typically lower than in CTV coverage, without a very clear trend in the differences (Table1).Except for intracranial treatments, the differences in the CTV due to the dose calculation algorithm MC dose were in general larger than the difference between the nominal plan and worst-case scenario after robustness evaluation (see example in fig. 1).

objectives were fulfilled the multi-criterial optimizer reduced the dose to all organs at risk until doses to PTVs were not affected. For each case, the automatic plan was compared with the approved plan manually optimized. The following dosimetric parameters were recorded: PTV D95, Parotid D mean , PRV Cord D 1cm3 , esophagus D mean . For each plan a modulation degree index (MD) was computed: MD= Total MU/[Sum of (Segment Area x Segment MU)/Total Beam Area].

Table1 Cost function used in automatic template. In brackets the exponents of the serial cost functions are shown Results The plans produced automatically fulfilled OAR constraints and reach objectives in 9/10 case. In the remaining case the plan produced automatically was not able to reach a minimum requirement of PTV coverage. Plan comparison is shown in table2. Automatic plans achieved lower dose to both parotids at the cost of a little reduction of dose to the high dose PTV and increase of dose to PRV cord. Modulation degrees of automatically produced and approved plans were not statistically different.

Dose Inde x

PB- MC(% )

PB(cGy(RBE) )

MC(cGy(RBE) )

Site ROI

Brai n

CTV

D99 5623

5601

0.4

H&N CTV

D99 5917

5592

6%

CSI PTVBrain D99 3456

3124

11

Table2 Comparison of automatic vs approved plans. Average value are shown, in bracket standard deviations are reported. Conclusion In this study the feasibility of the automatic plan generation using the multicriteria function of Monaco has been shown. EP-1848 Inaccuracies in proton dose calculation may be as significant as setup and range uncertainties M. Schwarz 1,2 , M. Innocenzi 1,3 , I. Giacomelli 1 , F. Fracchiolla 1 , V. Patera 3 , R. Righetto 1 1 Centro di Protonterapia, Protontherapy, Trento, Italy ; 2 TIFPA-INFN, Medical Physycs, Trento, Italy ; 3 Università degli Studi di Roma, Basic and Applied Sciences for Engineerin, Roma, Italy Purpose or Objective Plan robustness with respect to setup and range errors is considered a priority in protontherapy. However, geometrical uncertainties are not the only source of error in proton treatment planning. In this work we evaluate the effects of setup errors, range uncertainties and dose calculation algorithms on protontherapy dose distributions, to assess which uncertainties affect the most the difference between the nominal and actual dose distribution. Material and Methods 12 treatment plans used for clinical treatment at our centre were selected, covering four treatment sites: brain, head&neck, chordoma of the spine, and craniospinal axis. The plans, which were initially optimized with a PTV-based single field optimization technique and pencil beam (PB) dose algorithm, were

Brai n

Brainste m

D1 5770

5823

-1%

Mean dose 1993

H&N Lacrimal gland

9%

1829

586

2%

CSI Kidney Mean

dose 597

Conclusion In protontherapy, the effect of using two dose calculation algorithms may be larger than the effect of setup errors and range uncertainties, especially in CTV coverage for shallow targets. The minimum dose to the CTV is not guaranteed when plans are designed with PB, in particular for shallow targets. Ensuring plan robustness is important, but it should not come at the cost of less accurate dose calculation. With significant tissue heterogeneities and/or shallow