ESTRO 2021 Abstract Book

S775

ESTRO 2021

There is a need for robust and accurate segmentation algorithms for rectal tumor segmentation since manual segmentation of these tumors is susceptible to significant inter-observer variability. The deep learning-based segmentation algorithms proposed in this study are more efficient and achieved a higher agreement with our manual ground truth segmentations than a second expert annotator. Future studies will investigate how to train deep learning models on multiple ground truth annotations to prevent learning observer biases.

Poster discussions: Poster discussion 35: New delivery techniques

PD-0932 Scintillator characterization in an electron FLASH beam: first experiences V. Vanreusel 1,2 , A. Gasparini 1,2 , A. Giammanco 3 , M. Pacitti 4 , M. Cociorb 3 , E. D' Agostino 3 , G. Felici 4 , D. Verellen 1,2 1 Iridium netwerk, Medical Physics, Antwerp, Belgium; 2 University of Antwerp, Faculty of Medicine and Health Sciences, Antwerp, Belgium; 3 DoseVue N.V., Research and Development, Diepenbeek, Belgium; 4 Sordina IORT Technologies, Research and Development, Aprilia, Italy Purpose or Objective The recent re-introduction of the FLASH-effect, where a combination of ultra-high dose rates ( ∼ 40Gy/s) with very short treatment times ( ∼ 100ms) leads to a decrease of normal tissue radiotoxicity, opened a challenging field of research. The search for the biological processes behind the FLASH-effect and the clinical implementation of FLASH-RT requires reliable and accurate dosimeters. Scintillators are promising candidates for on-line dosimetry in FLASH-RT. A first characterization of the scintillator response in a highly tuneable electron FLASH beam is performed to validate its candidature. Materials and Methods Both the DoseWire™ Series 200 and an experimental variant of it , developed by DoseVue N.V. (Belgium), were irradiated using the commercial ElectronFlash4000 linac (SIT, Italy) that delivers electron FLASH beams with energies of 7 and 9 MeV. It allows independent adjustment of pulse length, pulse rate frequency (PRF) and number of pulses allowing to characterize the scintillators’ responses in varying FLASH conditions. First, the response with dose per pulse was investigated by 1) altering the pulse length between 0.5-4µs and 2) increasing the distance from the applicator exit window from 0 to 100cm (106-206cm from linac exit window). Next, the response to total dose was investigated by increasing the number of pulses from 20 to 500 pulses in the highest dose rate setting. Last, the system stability with PRF was investigated by varying the PRF from 5 to 245Hz in the highest dose rate setting. A circular 10cm diameter applicator was used for all irradiations. Results All scintillators show a high degree of linearity with pulse length (R² > 0.95), hence dose per pulse. Despite these high R²-values, a slight concave curvature can be observed for the measurements in contact. This is however not observed for measurements at a distance of 1m from applicator exit (i.e. at 206cm) window for the experimental system, suggesting saturation effects for measurements in contact. This effect was less evident when plotting response against SSD, where the results followed the inverse square law with R>0.95. All detectors show excellent linear response with increasing number of pulses, hence dose (R² > 0.99). But dose per pulse read-out shows some instability with PRF as it decreases with increasing PRF until it reaches a plateau at a PRF of 200 Hz. This might suggest some saturation due to electronic processing. Further investigation of MU stability and the dosimetric fiber material (PMMA) radiation induced opacity are currently ongoing. Conclusion We showed scintillators can deal with ultra-high doses per pulse with high linearity. By increasing the distance from the applicator exit window, saturation effects can be decreased, opening the possibility for correction factors and use in relative dosimetry. Optimization of the post processing, scintillator design and further characterization are required to validate the promising candidature of this technique for dosimetry in electron FLASH beams. 1 Helmholtz Zentrum München GmbH, German Research Center for Environmental Health, Institute of Radiation Medicine, Neuherberg, Germany; 2 Technical University of Munich, School of Medicine and Klinikum rechts der Isar, Department of Radiation Oncology, Munich, Germany; 3 Technical University of Munich, Physics Department, Garching, Germany Purpose or Objective Microbeam radiotherapy (MRT) is a novel treatment concept based on a micrometer-scaled spatial dose fractionation with high-dose peaks and low-dose valleys. In preclinical experiments, including lung tumor treatments, MRT has shown less severe side effects to normal tissue than conventional RT at the same tumor control rates. Prior to clinical trials, treatment planning studies with realistic clinical geometries are needed. We present a first MRT treatment planning study for a clinical lung tumor case. Dose to organs at risk (OAR) is hereafter reported as valley dose that primarily causes normal tissue damage. Materials and Methods As clinical reference, a conventional stereotactic RT plan for a peripheral lung tumor was created with the treatment planning system (TPS) Eclipse (Varian Medical Systems). The dose of 37.5 Gy in 3 fractions was prescribed to the 60% isodose and obtained by 9 coplanar beams (6 and 15 MV). The MRT plan was created with a hybrid algorithm combining Monte Carlo and convolution-based dose calculation for synchrotron x-rays (104 keV mean energy) [1], adapted for lung tissue and implemented into the open-source TPS 3DSlicer. Eight beam ports of the conventional plan were used for MRT with a peak width of 50 µm and a peak-to-peak distance of 400 µm. Neighboring MRT ports were shifted by 50 µm for a homogeneous target dose by an PD-0933 Microbeam radiotherapy planning for a clinical lung tumor case J. Winter 1,2,3 , K.M. Kraus 2 , M. Ahmed 1,2,3 , S.E. Combs 1,2 , J.J. Wilkens 2,3 , S. Bartzsch 1,2