ESTRO 2020 Abstract book

S841 ESTRO 2020

better dose conformity. AP plans are preferred for slightly superior dose falloff. PO-1475 Automated planning for pre-selection of head&neck patients for proton therapy J. Kouwenberg 1 , J. Penninkhof 1,2 , S. Habraken 1,2 , J. Zindler 1,2 , B. Heijmen 1 , M. Hoogeman 1,2 1 Erasmus Medical Center, Radiation Oncology, Rotterdam, The Netherlands ; 2 HollandPTC, Radiation Oncology, Delft, The Netherlands Purpose or Objective A comparative proton therapy (PT) treatment plan made at a PT center is often mandatory to justify the choice of PT for an individual patient. In practice, a clinician selects patients to be sent to a PT center for a comparative PT plan based on clinical factors, images, and treatment characteristics. However, this selection is subjective and dosimetric information is not included in the decision making process. Moreover, the PT treatment planning is labor-intensive and time-consuming. Therefore, it may withhold PT from patients that would have benefitted, or conversely could result in an unnecessary work and delay for patients for whom the comparison turns out to be negative. To overcome these undesired scenarios, we developed a novel automated procedure for selecting head & neck cancer patients for a comparative PT plan made at a PT center, using a non-clinical automated IMPT planning system (Erasmus-iCycle). Material and Methods Erasmus-iCycle was commissioned for a Varian ProBeam pencil beam scanning system used in the PT center. The wish-list, used to control the prioritization in the automated treatment planning process, and the robust optimization settings were configured to mimic the plans generated at the PT center. Subsequently, 18 H&N patients, who were referred to the PT center for a comparative PT plan, were selected. Differences in OAR doses and corresponding normal tissue complication probabilities (NTCP) between Erasmus-iCycle and the PT plans were determined. To this end, we evaluated NTCP models used in the Dutch model-based approach for PT patient selection, which are xerostomia, dysphagia, and tube feeding dependency. Results After commissioning the differences in range, spot size (70 – 244 MeV) and range-shifted spot size (100 – 244 MeV) were within 1, 0.2, and 0.7 mm, respectively. Comparing the Erasmus-iCycle and the PT center plans, the mean (± 1SD) differences in contralateral parotid gland, oral cavity, PCM superior, PCM inferior, and crycopharyngeus doses were 0.6 (± 2.4) Gy, -0.5 (± 3.6) Gy, 1.6 (± 3.7) Gy, -1.0 (± 4.3) Gy, and -2.37 (± 5.53) Gy, respectively. These OAR dose differences led to mean (± 1SD) differences in the NTCP for xerostomia, dysphagia, and feeding tube dependency -0.7 (± 2.7)%, -0.6 (± 2.3)%, and -0.6 (± 1.8)%, respectively. The outliers could be explained by a more conformal dose distribution that was achieved with Erasmus-iCycle, and the fact that the PT center sometimes allowed local underdosing of the robust target in order to spare OAR’s. Conclusion It was possible to approximate PT plans, which were manually generated at a PT center, with a non-clinical automated treatment planning system. This paves the way for an objective and efficient selection of patients who qualify for a comparative PT treatment plan, for example as part of the Dutch model-based approach.

on MRI T1 sequence). A 2 mm isotropic margin was added to obtain the boost target volume (BTV). The PTV was defined as the entire vertebra containing the lesion, with an isotropic 2 mm margin. Doses of 30 Gy to BTV and 21 Gy to PTV were simultaneously prescribed in three fractions. All patients were planned with dual arc VMAT technique using 6 MV FFF beams. For AP plans, Autoplanning used a progressive optimization algorithm to continually adjust initial targets/OARs objectives. Tuning structures and objectives are automatically added during optimization to increase the dose fall-off outside targets and improve the dose conformity. Optimal coverage for BTV and PTV was considered D95% ˃ 95% of each prescription dose. Compliance of OARs constraints was considered a priority: spinal cord: Dmax<22 Gy and V18Gy<0.1 cc; spinal canal: V18Gy<0.1 cc. Dose statistics for target coverage, OARs sparing, conformity indexes (CI) and the R50% (ratio between volume receiving 50% of the prescribed dose and PTV volume) were compared. A Wilcoxon paired-test was performed for plan comparison (p < 0.05 as statistical significance). Results Both AP and MP plans were considered clinically acceptable. A summary of dosimetric results is reported in Table 1.

Respecting all OARs constraints, AP and MP plans provided similar target coverage for PTV D95% (MP: 96.4 ± 2.2%, AP: 97.5 ± 4.8%, p = 0.173) but significant improvement for BTV (MP: 94.7 ± 7.2%, AP: 96.2 ± 8.8%, p = 0.046). AP plans exhibited higher conformity compared to MP plans for both BTV (MP: 2.4 ± 0.9%, AP: 2.3 ± 0.8%, p = 0.046) and PTV (MP: 2.5 ± 0.9%, AP: 2.3 ± 0.6%, p = 0.046), respectively. R50% was significantly better for AP plans (MP: 9.1 ± 2.0%, AP: 7.7 ± 1.3%, p = 0.028). Skin dose was significantly lower with AP plans. Last, planning time was reduced to about 1 hour for AP plans. Figure 1 shows the dose distributions for a representative patient.

Conclusion Automated SBRT-VMAT planning for complex spinal dose distributions is feasible. Pinnacle3 Autoplanning reported