ESTRO 2021 Abstract Book

S454

ESTRO 2021

how much data is actually required for optimal performance. Also, current work in deep learning has focussed on generating dose distributions only, not deliverable treatment plans.

Symposium: Image guidance in particle therapy: From state-of-the-art towards real-time ART

SP-0581 Patterns of practice and guideline development: A status update from the EPTN image guidance working group A. Bolsi 1 , A. Hoffmann 2 1 Paul Scherrer Institute, Centre for Proton Therapy, Villigen, Switzerland; 2 OncoRay, National Center for Radiation Research in Oncology, Medical Radiation Physics, Dresden, Germany Abstract Text Image guidance for particle therapy (IGPT) is essential to guarantee accurate dose delivery and to minimize the effect of range uncertainties related to patient misalignment and anatomical changes. Multiple imaging modalities are available for clinical IGPT such as daily orthogonal kilovoltage (kV) X-ray imaging, cone-beam CT (CBCT) and in some centers also in-room CT imaging, implemented both for patient positioning and for offline monitoring of anatomy changes. Given the lack of standardized procedures for IGPT, most PT centers have developed their own strategies, which are based on their available infrastructure, technical implementation and dose delivery strategy. The need to collaborate among centers has been recognized with two main primary goals: a) to develop clinical consensus guidelines for IGPT; and b) to define and pursue research projects to optimize existing IGPT techniques and to develop new ones. The working group on “ Image Guidance in Particle Therapy ” of ESTRO’s EPTN Task Force, has identified, as the first step in the development of clinical guidelines, a detailed description of the current clinical practices in the European PT centers. In order to achieve this, the sub-working groups defined in 2017, have prepared multiple detailed surveys on IGPT for the specific localizations such as brain and head and neck, breast, abdomen and pelvis, craniospinal axis and extremities. These surveys have been sent out to all centers in 2019/2020 and responses have been collected. The analysis of the response data has started with the participation of the sub-working groups. Two additional surveys for IGPT in the thorax region, considering also 4D treatments, are under finalization to be sent out soon to the participating centers. The results of the body-site specific surveys will be included in the clinical consensus guidelines. During the 2020 meeting of our working group, multiple areas of research and collaborative projects have been identified and discussed with the sub- working group coordinators. It was decided to focus on a multicentric evaluation of the inter- and intrafraction patient positioning accuracy for brain tumor patients undergoing PT. Results for IGPT should be analyzed for all the participating centers, considering different positioning and image guidance procedures together with the immobilization devices. In this contribution, an overview of the available results of the body-site specific surveys will be presented, together with the roadmap for the next activities of the IGPT working group of the EPTN. Abstract Text Cone-beam computed tomography (CBCT) is playing an increasingly important role as in-room imaging modality for proton therapy. A range of proton-specific devices mounted at the treatment gantry, the patient table or the ceiling have been introduced in clinical proton facilities. Up to now, usage of these devices is limited for accurate pre-treatment patient alignment, even though proton therapy might greatly benefit from daily CBCT based treatment adaptation. This restricted use of CBCT is related to the fact that, despite accurately capturing the patient anatomy, image quality is considerably degraded with respect to diagnostic quality CT, mainly due to the detection of scattered photons. In particular, image quality and CT number accuracy are not sufficient to perform accurate proton dose calculations for treatment adaptation [1]. However, over the last years, a variety of correction methods aiming at rendering the in-room CBCT images suitable for accurate proton dose calculation have been suggested in the literature and will be presented in this contribution. Approaches range from simple population based CT number rescaling, over Monte-Carlo- based scatter estimation and correction, to deformable image registration (DIR) based methods. The latter either exploit images obtained from CT-to-CBCT DIR directly or use the generated virtual CTs to perform projection-based intensity correction. In general, all of the previously mentioned methods suffer from either limited accuracy (at least for certain treatment sites) and/or excessive CBCT correction times. This renders them unsuitable in the scope of online CBCT based adaptive proton therapy, where correction times in the order of few seconds and high CT number accuracy are crucial. Consequently, there is recently a strongly growing scientific interest in performing CBCT correction by means of deep convolutional neural networks (CNN), which promise both accurate and fast results. The most common deep learning approaches feature U- shaped CNNs or so-called generative adversarial networks (GANs) [2]. Besides deep learning, another area of current research is the extension of CBCT correction methods to treatment sites affected by intra-fractional organ motion. The generated corrected day-of-treatment 4DCBCTs might not only enable accurate dose calculation and treatment adaptation, but also estimation of the actually applied treatment dose under consideration of interplay effects between dose delivery and target motion [3]. In this contribution, strengths, weaknesses and future challenges for the clinical introduction of conventional and deep learning based correction methods, paving the way towards CBCT based online adaptive proton therapy, will be discussed. SP-0582 Online CBCT-based proton range and dose calculation C. Kurz 1 1 LMU Munich, Department of Radiation Oncology, Munich, Germany

[1] Landry et al., 2018, Med. Phys. 45(11), e1086-e1095 [2] Giacometti et al., 2020, Phys. Med. 76, 243-276