ESTRO 2020 Abstract book

S741 ESTRO 2020

PO-1314 Creation of a 3D-printed plastic mouse phantom for pre-clinical dosimetry and quality assurance S. Kampfer 1,2 , S. Dobiasch 1,3 , S.E. Combs 1,3,4 , J.J. Wilkens 1,2 1 School of Medicine and Klinikum rechts der Isar- Technical University of Munich, Department of Radiation Oncology, Munich, Germany ; 2 Technical University of Munich, Physics Department, Garching, Germany ; 3 Helmholtz Zentrum München, Institute of Radiation Medicine IRM, Neuherberg, Germany ; 4 Helmholtz Zentrum München, Institute for Innovative Radiotherapy iRT, Neuherberg, Germany Purpose or Objective The increasing use of high-precision small animal irradiation devices results in a correspondingly high demand for dosimetry and quality assurance (QA) on such machines. Some demands are easily manageable with standard measurement equipment. Nevertheless, some procedures require non standardized equipment, such as special phantoms. We tested an approach to print an individual mouse phantom on a common tabletop 3D printer and characterized it by absolute dose measurements. The purpose of this work is to present a fast and simple way to produce individual mouse phantoms with adequate dosimetric properties. Material and Methods To create a 3D model of the mouse phantom, we used a cone-beam CT scan of a real mouse performed on our Small Animal Radiation Research Platform (SARRP, Xstrahl Ltd., Camberley, UK). We contoured the body of the mouse on the CT scan in the open source software 3DSlicer (version 4.10.2) and exported the model to a 3D printer software (Dremel DigiLab 3D Slicer, version 1.2.2, Dremel, Racine, WI, USA). The mouse was then solid 3D-printed on a Dremel 3D45 printer from polyactid acid (PLA, C 3 H 4 O 2, Dremel). We tested the dosimetric properties of the PLA compound in the energy range used in small animal radiotherapy (up to 220 keV) by comparing simple depth dose measurements with expected results in soft tissue. The 3D printed mouse was validated by measuring the exit dose behind the phantom and behind a real mouse with radiochromic film (Gafchromic EBT3, Ashland, USA). The body diameter was 27±1 mm in both cases. The used beam configuration on our SARRP was 220 kV, 13.0 mA, broad focus, 0.15 mm copper filter, 3x3 mm² fixed collimator and an irradiation time of 68 seconds. Results Creation and printing of the mouse phantom was straightforward and took about four hours for medium quality. The material PLA was used due to good availability and easy usage (biodegradable, less likely to warp and no smell during printing). The linear attenuation coefficient (in the interesting energy range) is very similar to that of soft tissue. For 222 keV (58.5 keV) it is 0.145 1/cm² (0.237 1/cm²) for PLA versus 0.125 1/cm² (0.210 1/cm²) for soft tissue. Depth dose measurements showed excellent agreement with the expected gradient in soft tissue with less than 1% difference of dose in depths of 1.5 mm and 10 mm. The exit dose behind the phantom was determined to be 1.51 Gy, whereas the exit dose behind the live mouse was 1.54 Gy for the same irradiation parameters (difference of only about 2%). Conclusion With PLA as print material we could produce a mouse phantom in a reasonable time that gives very good results of the measured dosimetric parameters. We showed that the exit dose after the phantom is nearly identical to that after a real mouse. The results promise a broad range of possible applications like end-to-end tests and measurements with a detector or film inside of an individualized phantom.

United Kingdom ; 3 University College London Hospital, Department of Oncology & Radiotherapy Department, London, United Kingdom Purpose or Objective Virtual phantoms are three-dimensional surrogates of the anatomy of the human body. Detailed age-specific phantoms are usually built from healthy individuals, therefore biased and not truly representative of the cancer population. To address these limitations, we investigate the feasibility of a population-based approach to automatically generate virtual voxelized phantoms representative of childhood cancer patients across different ages. These have important applications in risk assessment of novel therapeutic modalities and reconstruction of historical radiotherapy (RT) doses. Material and Methods We combine the anatomical information from a group of RT paediatric patients to build population-based average models (virtual phantoms), a methodology based on groupwise deformable image registration (DIR). Groupwise image registration is a process of spatial normalisation that iteratively alternates between pairwise registration of all subjects to a reference image (the “average” anatomy) and updating this reference image as the mean of the resulting pairwise registrations. To generate age-specific anatomical models, we use an open-source DIR algorithm (NiftyReg), image processing pipelines developed and optimised in-house, a database of 58 paediatric CTs and available segmentations (median age 6.8 y, range: 1.2– 15.8 y). The target age of the output model is defined during the algorithm initialisation. Twenty subjects (closest to the initial target age and anatomy) are then automatically selected and used to build the virtual CT. To generate the delineations all data clinically available is used. The final output is a virtual CT-like image (and corresponding segmentations) representative of the subgroup used to generate it. Results To test our implementation, we built nine voxelized virtual phantoms for ages ranging from 4.1 to 13.4 y. Models were not sex-specific. To investigate the anatomical plausibility of the models, we compared the volume, shape and pixel intensity of key organs-at-risk against values measured for individual patients. The figure shows the volume of key organs for the virtual phantoms against the population data. The relative difference in volume between the virtual models and a linear fit of the population data was 7.2±5.7 %, 7.8±3.5 %, 6.9±4.7 %, 2.2±1.1 %, and 7.8±6.4 % for lungs, kidneys, globes, brain and liver, respectively. In terms of organ height, the relative differences were 2.5±1.5 %, 2.0±1.6 %, 3.9±3.1 %, 2.7±1.0 %, and 7.3±4.0 %. The absolute differences in pixel intensity were 23±12, 6.0±3.0, 2.7±0.6, 3.9±0.3 and 8.1±1.6 HU.

Conclusion We found promising evidence that population-based average models are representative of paediatric patients undergoing external RT at various ages, particularly in the head and thorax. Further development and evaluation studies are required, namely developing sex-specific phantoms, evaluate performance on independent datasets and investigate the anatomical plausibility for other organs-at-risk in the abdomen and pelvis.