ESTRO 2020 Abstract book

S461 ESTRO 2020

SP-0749 A systematic, large-scale planning comparison for patient selection in proton therapy S. Habraken 1,2 1 Erasmus Medical Center Cancer Institute, Radiation Oncology, Rotterdam, The Netherlands ; 2 holland Proton Therapy Center, Radiation Oncology, Delft, The Netherlands Abstract text In the Netherlands, the indication for proton therapy is model-based in a majority of cases. A photon treatment plan from the referring institute is compared to a proton plan on the same dataset (planning CT scan and delineation), made by the proton therapy facility. Differences in normal tissue complication probability (NTCP) are calculated and evaluated. For head-and-neck cancer, proton therapy is reimbursed if the model risk of xerostomia, dysphagia, and tube feeding dependence can be reduced by 10%, 10%, and 5%, respectively. Proton therapy is indicated for neurological tumors with a favorable prognosis (10-year survival 50% or more) if the mean dose to the supratentorial brain outside the target and/or the hippocampi can be reduced by 5% or more. For breast cancer, proton therapy is reimbursed if the model risk of (late) heart toxicity can be reduced by 2% or more. In this talk, I will review and discuss clinical experience with large-scale plan comparisons for head-and-neck, neurological and breast cases. The following topics will be addressed: (i) the fraction of patients that is referred for each treatment (sub)site after a plan comparison, challenges and optimization of the workflow in clinical practice, NTCP differences between the plan-comparison plan and the clinical treatment plan, and NTCP stability during treatment (plan adaptation). Finally, I will address some future developments, including plan comparison based on fully automated treatment planning or dose/NTCP prediction with machine learning. SP-0750 Problems solved and open questions in clinical practice of proton treatment planning. M. Schwarz 1,2 1 S. Chiara Hospital, Protontherapy, Trento, Italy ; 2 tifpa- Infn, Medical Physics, Trento, Italy Abstract text The recent past of proton treatment planning in clinical practice saw significant developments, as well as the beginning of new issues likely to be solved in the near future. Pencil beam scanning is the indisputed standard for most proton treatments, and this allowed large scale implementation of techniques based on either single field (SFO/SFUD) or multi-field (MFO/IMPT) optimization, showing how MFO is able in several cases to produce plans that are more robust, not less, than the SFO (and, even more, passive scattering) counterpart. The field of robustness optimization shifted its focus from algorithms development to the introduction of these approaches in clinical practice, where it is now in a phase of consolidation, where standardization of practices is both needed and highly sought by many centers. Plan robustness evaluation, which hasn't been the subject of research nearly as much as robust optimization, raises interesting questions that go beyond the practice of proton therapy and show the need of more efforts to tackle an issue related to photon therapy too. Dose calculation was associated with both bad news (e.g. the inadequacy of some dose calculation algorithms in challenging situations) and good news, such as the

introduction of Monte Carlo dose calculation in routine planning. We can now say that dose calculation accuracy is not the weak link in the overall accuracy of proton therapy. On two aspects proton planning is seemingly running behind photon techniques: (online) adaptive planning and autoplanning. Fast adaptive planning is still far from being the norm, and at the same time large scale availability of high quality daily imaging and the use of hypofractionation regimes are still lacking. Albeit it is difficult to identify which are the causes and which are the consequences, demonstrating the feasibility in clinical practice of efficient replanning approaches is a priority. Protons have specific needs, such as model-based patient selection, which would greatly benefit from autoplanning. Instead, autoplanning for proton therapy is in its infancy. Methods for automatic beam selection are probably a "low hanging fruit" worth our attention to enable large scale adoption of autoplanning in the near future. SP-0751 3D bolus: practical aspects S. Goncalves 1 1 Instituto Português de Oncologia do Porto Francisco Gentil- EPE, Física Médica, Porto, Portugal Abstract text In radiation therapy, the need to irradiate superficial lesions is common; these lesions are often at shallow depth or on the skin of the patient. However, one of the characteristics of the radiation used in radiotherapy treatments is that the maximum dose deposited lies at a given depth, with skin-sparing property. Therefore, a bolus, a natural or synthetically developed material, is placed on the surface area to be irradiated and acts as a tissue layer to provide a more effective treatment in the superficial lesions. Nevertheless, it is difficult for the commercially available flat-form boluses to make full contact with irregularly shaped patient skin, as it occurs in post-mastectomy chest wall, causing the presence of air gaps between the bolus layer and the patient's skin that may influence the planned dose distribution. In fact, this phenomenon is not considered by TPS and may cause serious changes between the planned dose distribution and the actual dose administered during the treatment. The decrease of the air gaps will give a better conformation of the dose distribution to the target volume, improving the accuracy of the radiation treatment. The use of 3D printing techniques to create a patient specific bolus facilitates correspondence with the patient skin, yielding agreement between the planned and delivered doses. This lecture aims to describe dosimetric properties and practical aspects of flexible materials for 3D printed bolus and to show how to produce efficient and more comfortable customized bolus for patients. Symposium: Innovations in radiotherapy