ESTRO 37 Abstract book

S108

ESTRO 37

Purpose or Objective In the Netherlands, the model-based approach is introduced to select patients for proton therapy using normal tissue complication probability (NTCP) models. Patient selection is based on ΔNTCP values derived from a proton and a photon dose distribution. For planning efficiency, fully automated treatment planning is required. In this study, we developed a dose mimicking and reduce (DMR) algorithm to automatically generate a robust multi-field optimized proton plan given a reference photon dose distribution and target and organs at risk (OAR) delineations. The DMR plans were compared against manually optimized (reference) robust proton plans. Material and Methods The DMR algorithm was evaluated using clinically accepted photon dose distributions of five head and neck cancer patients. The first step of the DMR algorithm comprised of DVH-based mimicking of the photon dose distribution in the clinical target volumes (CTVs) and OARs, accounting for setup and range uncertainties of 5.0 mm and ±3%, respectively. Plan robustness was achieved by mimicking the nominal photon dose in 21 perturbed scenarios. The use of protons provides leeway to improve the dose to the non-target tissues. The second step of the optimization therefore aims to reduce OAR doses as much as possible while retaining the target coverage and uniformity achieved in the first step. We evaluated each DMR plan against the photon plan and reference proton plan in terms of plan robustness and NTCPs of xerostomia, dysphagia and tube feeding dependence. Plans were considered sufficiently robust if CTV V95%≥98% in the voxel-wise minimum dose, which was derived from 14 perturbed dose calculations with 5.0 mm shifts and ±3% range uncertainty (28 scenarios in total). To avoid any bias from proton beam angle selection the same beam configuration was used in the DMR plans and the reference proton plans. Results The quality of the DMR proton plans was very similar to the reference proton plans in terms of dose values and robustness. All proton plans showed a primary and elective CTV coverage in the voxel-wise minimum dose of V95%>98%. Figure 1 illustrates a transversal view of a photon, DMR and reference proton plan. Within the studied cases, the sum of the ΔNTCPs (difference between photon and proton plans) ranged from 11.4 – 42.0% and 4/5 patients were selected for proton therapy based on either the DMR or reference proton plan. The mean dose difference between the DMR and reference proton plans of the parotid glands, swallowing muscles, and supraglottic larynx ranged from -4.1 to 4.5 Gy. This translated into differences in ΔNTCPs of -1.0 to 2.8%.

head and neck cancer patients for proton therapy by combining the DMR algorithm with the model-based approach. PV-0202 Towards a more robust 4D optimization for PBS-based moving tumor treatments Y. Zhang 1 , I. Huth 2 , D.C. Weber 1 , A.J. Lomax 1 1 Paul Scherrer Institut, Centre for Proton Therapy, Villigen-PSI, Switzerland 2 Varian Medical Systems – Particle Therapy GmbH, Particle Therapy GmbH, Troisdorf, Germany Purpose or Objective Complementary to classic motion mitigation approaches (rescanning, gating, breath hold, tracking or combinations of these), 4D optimization can inherently include motion into the process of proton pencil beam weight optimization, resulting in a ‘motion-robust’ plan. We aim here to evaluate this approach systematically using 4DCT-MRI datasets, and to investigate the possibility of increasing plan robustness by incorporating rescanning into the optimization. Material and Methods Using nine 4DCT(MRI) datasets of liver tumors (CTV:100- 400cc), 3D plans (Fig. 1) were generated on the end-of- exhalation CT phase using a geometric ITV (encapsulating CTVs at all 4DCT phases), and a 4D dose calculation (Fig. 1) was performed by considering regular and irregular motion patterns (range: 8-20mm; period: 3.3-6.3s) with and without rescanning (x5 layered). In addition, 4D, directly optimized plans, which accurately take into account beam delivery dynamics and organ motion, were calculated to the CTV (on reference CT) without expanded internal margins. The robustness of optimized 4D plans were assessed by recalculating the optimized 4D plans under variable motion scenarios with phase delays up to 1s (in 50ms intervals). All re-calculated 4D plans were quantified and compared using homogeneity index (HI) of D5-D95 in the CTV as well as V10%, V20% and V60% in the healthy liver (liver-CTV).

Conclusion The DMR algorithm automatically optimized a robust proton plan using a photon reference dose. For the studied cases, the DMR plans qualified the same patients for proton therapy as the manually optimized proton plans. We developed a fully automated tool to select

Results Independent on the motion pattern (regular or irregular), amplitudes (up to 20mm) or periods (either fast or slow breathers), interplay effects could be effectively mitigated using 4D optimization alone, with HI in the CTV