ESTRO 36 Abstract Book

S831 ESTRO 36 _______________________________________________________________________________________________

Physics, Montreal, Canada 3 McGill University Health Center - Glen Site, Radiation Oncology, Montreal, Canada Purpose or Objective Despite considerable increase in the number of degrees of freedom with recent radiotherapy optimisation algorithms, treatments are typically delivered using only a single modality. Mixed-modality plans such as electron- photon provide dosimetric advantages for sites with a superficial component. The column generation method, which is an iterative method that finds the aperture with the largest potential to improve the cost function at every iteration, is well suited for mixed-modality optimisation as the aperture generation and modality selection problem can be solved quickly. We assess the performance of the column generation method applied to mixed- modality planning and investigate its behaviour under different modality mixing schemes. Material and Methods The column generation method was applied to a chest wall case (Fig. 1). Photon beamlets were created for a coplanar distribution of beam angles every 20º around the patient. In addition, 5 shortened-SAD (70 cm) electron beam angles were included and beamlets were generated for energies of 6, 9, 12, 16, 20 MeV. A photon MLC acted as the sole collimating device for electrons.

therapy (VMAT) plans, and M IMPT , consisting of 40 intensity modulated proton therapy (IMPT) plans. Modeled OARs included the parotid glands (PGs), submandibular glands (SMGs), oral cavity (OC) and individual swallowing muscles. For 10 evaluation patients, PSP was used to derive mean predicted OAR doses generated by M VMAT /M IMPT , which were further interpreted using NTCP models for the following OARs: 1: PGs (stimulated flow ratio<25% pre-treatment, Dijkema IJROBP 2010), 2: SMGs (grade 4 toxicity, Murdoch-Kinch IJROBP 2008), 3: OC (grade 3 mucositis, Bhide R&O 2012) and 4: Superior pharyngeal constrictor muscle/supraglottic larynx (sPCM/SGL, grade 2-4 dysphagia, Christianen R&O 2012). Results Figure 1 shows a screenshot of a DVH comparison for the left PG using PSP , with the red/blue shaded regions representing the dose predictions by M VMAT /M IMPT . On average, mean CL/IL parotid gland doses were 21.3±3.3/33.6±4.2Gy and 14.5±6.0/28.8±5.0Gy using M VMAT and M IMPT , respectively, associated with a 5.0/7.8% reduction in NTCP. Conversely, predicted CL SMG doses were 33.0/31.9Gy with M VMAT /M IMPT , with a 0.3% difference in estimated NTCP. The largest reduction using M IMPT was noted for the occurrence of oral mucositis (15.6%), with an average OC mean dose reduction of 16.9Gy. Finally, a 4.5% decrease in dysphagia was predicted using IMPT, associated with sPCM/SGL dose reductions of 6.1/5.0Gy.

Photon-only (IMRT), electron-only (MERT) and mixed electron-photon (MBRT) treatment plans were created using the same optimisation constraints. To analyse the sensitivity of treatment plans to initial conditions, a perturbation on the original mixed-modality treatment plan was created by forcing the first 50 apertures to be photon apertures before allowing other modalities. Finally, the efficiency and plan quality of four different modality mixing schemes was analysed by creating treatment plans with more than a single aperture per iteration of the column generation loop. Results The MBRT plan produced better target coverage and homogeneity while preserving the normal tissue-sparing advantages of electron therapy (fig. 2, inset), with a final cost function between 25-30% of the values for IMRT and MERT. The fraction of total dose among modalities in the treatment plan with the perturbation (fig. 2, full lines) converged to the unperturbed treatment plan (fig. 2, dashed lines) with identical plan quality.

Conclusion By including standard dosimetric and NTCP metrics, PSP may assist in optimal treatment selection for individual patients. This analysis is based on a DVH line that is predicted by RapidPlan, without requiring the actual (time consuming) creation of treatment plans. This makes the plan comparison process efficient and transparent. Note that the proton versus photon comparison for HNC was solely used as a paradigm, this study was not intended to investigate the accuracy of RapidPlan for protons. EP-1545 On mixed-modality radiation therapy optimization using the column generation approach M.A. Renaud 1 , M. Serban 2 , C. Lambert 3 , J. Seuntjens 1 1 McGill University - Downtown Campus Department of Oncology, Physics, Montreal, Canada 2 McGill University Health Center - Glen Site, Medical