A comparison of two pencil beam scanning treatment planning systems for proton therapy

Abstract Objective Analytical dose calculation algorithms for Eclipse and Raystation treatment planning systems (TPS), as well as a Raystation Monte Carlo model are compared to corresponding measured point doses. Method The TPS were modeled with the same beam data acquired during commissioning. Thirty‐five typical plans were made with each planning system, 31 without range shifter and four with a 5 cm range shifter. Point doses in these planes were compared to measured doses. Results The mean percentage difference for all plans between Raystation and Eclipse were 1.51 ± 1.99%. The mean percentage difference for all plans between TPS models and measured values are −2.06 ± 1.48% for Raystation pencil beam (PB), −0.59 ± 1.71% for Eclipse and −1.69 ± 1.11% for Raystation monte carlo (MC). The distribution for the patient plans were similar for Eclipse and Raystation MC with a P‐value of 0.59 for a two tailed unpaired t‐test and significantly different from the Raystation PB model with P = 0.0013 between Raystation MC and PB. All three models faired markedly better if plans with a 5 cm range shifter were ignored. Plan comparisons with a 5 cm range shifter give differences between Raystation and Eclipse of 3.77 ± 1.82%. The mean percentage difference for 5 cm range shifter plans between TPS models and measured values are −3.89 ± 2.79% for Raystation PB, −0.25 ± 3.85% for Eclipse and 1.55 ± 1.95% for Raystation MC. Conclusion Both Eclipse and Raystation PB TPS are not always accurate within ±3% for a 5 cm range shifters or for small targets. This was improved with the Raystation MC model. The point dose calculations of Eclipse, Raystation PB, and Raystation MC compare within ±3% to measured doses for the other scenarios tested.


| INTRODUCTION
Many proton centers around the world are implementing pencil beam spot scanning as a preferred delivery method for proton therapy. Spot scanning beams have the ability to modulate energy as well as intensity without the use of apertures/collimators and compensators in the beamline. 1 At our center spot scanning techniques are exclusively used. When considering how to most effectively treat patients with proton therapy, different treatment planning system provide different advantages and disadvantages. 2,3 These differences in different treatment planning systems were discussed in various articles for photons [4][5][6][7][8] and for protons. [9][10][11] Pencil beam scanning can also be simulated with Monte Carlo models, 12 although this is more time consuming and not practical for everyday treatment planning.
At our center we use two commercially available clinical treatment planning systems (TPS), Eclipse and Raystation. In this work, we compare the analytical dose calculation algorithms for these systems, as well as for a Raystation Monte Carlo model to corresponding measured point doses. The main issues using analytical algorithms for range shifters are that the average proton energy decreases when scattering angle increases during the non-elastic scattering processes as shown in Lin et al., 12 especially for small fields or fields including a range shifter. 10,11 We therefore compare the planning systems to quantify these inaccuracies.  [14][15][16][17][18] The TPS were modeled with the same beam data acquired during commissioning. 19,20 To commission the TPS the integrated depth dose curves (IDDs) and spot profiles in air were measured over the entire range of energies available, starting from 245 MeV, 240 MeV, and then every 10 MeV down to 70 MeV. For both TPS only the spot profiles at one gantry angle could be used for commissioning, even though there is a change in full width half maximum of the spots as function of gantry angle. 20 We choose to use the spot profiles measured at gantry 0 every 10 MeV from 70 MeV to 245 MeV, and verified that the variation at other angles and among all treatment rooms are within 15%. In addition to the spot profiles in air at isocenter, they were also measured for both TPS at 10 cm and 20 cm above and below isocenter as well as for each range shifter commissioned. Range shifters provide dose coverage for more superficial tumors. For both TPS the 5.7 cm, 3.42 cm and 2.28 cm water equivalent thickness (WET) range shifters were commissioned. The range shifters were commissioned with an airgap of 26 cm. This distance is an approximation of what will be used clinically for most cases. Smaller airgaps will be used during planning, but might not always be possible because of patient couch collisions as the snout is extended.

| MATERIALS AND METHODS
Absolute calibration was done using the TRS398 protocol. 21 For Eclipse the absolute dose measurement was done at a water equivalent depth of 2 cm for all the energies measured for the IDDs.
Raystation required measurements of absolute dose at a water equivalent depth at the position between 1 cm depth and half way between maximum of the Bragg peak for the same energies. To verify the modeled TPS measurements, 35 typical plans were made with the TPS, 31 without a range shifter and four with a 5 cm range shifter (5.7 cm WET). A 40 cm water equivalent virtual phantom was created with the isocenter placed at 20 cm depth. The Hounsfield units for this virtual phantom was forced to 0 for Eclipse, resulting in a mass density of 1.024 g/cm 3 and a relative stopping power of 1.002. For Raystation the same CT calibration curve was used and the phantom material was set as water, resulting in a mass density of 1.00 g/cm 3 . First generic water equivalent phantom plans were created for targets close to the surface, deeper in the phantom and at a midrange. Small targets (2 9 2 9 2 cm 3 ) and large targets   range shifter is small, two of the four plans had percentage differences in more than 3% and the measured doses are mostly higher than the TPS doses. For Raystation PB plans the measured doses are lower than the TPS doses for all the plans and 3 of the 4 plans have percentage differences higher than 3%, with a maximum of 7.16%. Comparing the TPS doses, the Raystation PB plans with a 5 cm range shifter give doses higher than those calculated by Eclipse by more than 3% for 3 of the 4 plans. This can in part be due to the slightly lower mass density of 1.00 g/cm 3 assigned for the water phantom in Raystation, compared to 1.024 g/cm 3 in Eclipse (for HU of 0), resulting in a difference in stopping power in water compared to Eclipse.

| RESULTS
In Fig. 3 and Table 2