ESTRO 2021 Abstract Book

S666

ESTRO 2021

Results Our abdominal phantom presented highly anthropomorphic contrasts in MRI and CT with tissue-equivalent T1, T2 and HU values for the liver and the kidneys. Furthermore, the anthropomorphic phantom was able to perform realistic and reproducible real-time breathing motion. For this, extracted diaphragm movement from 4D MRI data was translated to the phantom in real-time indicating that the liver and the kidney movement of the phantom was precisely timed and resulted in a neglegteable deviation in amplitude of 0.2% and 2.2% for liver and kidney, respectively compared to reference patient. Conclusion Together, results demonstrated the suitability of this anthropomorphic phantom for MR-Linac imaging and end-to-end tests. Multimodality was shown by testing the phantom in MRI and CT. Furthermore, liver movement agreed well with measured patient data. To conclude, this multimodal abdominal phantom incorporates realistic diaphragmatic motion which may offer a method to test novel treatment options. PD-0830 GPD1 high expression in glioblastoma stem cells confers radiation resistance C. Xu 1 , M. Bonrouhi 1 , M. Roscher 1 , M.J. Besso 1 , R. Lange 1 , W.W. Hadiwikarta 1 , H. Liu 2 , I. Kurth 1 , M. Baumann 1 1 German Cancer Research Center (DKFZ), Radiotherapy and Radiation Oncology, Heidelberg, Germany; 2 German Cancer Research Center (DKFZ), Molecular Neurogenetics, Heidelberg, Germany Purpose or Objective Glioblastoma (GBM) is the most common primary brain tumor with almost 100% recurrence rate after surgery and radiochemotherapy. Previously, the Liu laboratory at DKFZ found that glycerol-3-phosphate dehydrogenase 1 (GPD1) was expressed specifically in brain tumor stem cells (BTSCs), and that GPD1 expression drove tumor relapse after chemotherapy. So far the role of GPD1 expression in GBM receiving radiation therapy remains unclear. We therefore investigated the role of GPD1 expression in GBM receiving radiation therapy with or without chemotherapy. Materials and Methods We utilized 4 primary mice brain tumor stem cell lines (BTSC-tum1, BTSC-tum6, BTSC-A, and BTSC-Azer), and conducted the cell spheres’ growth delay assay and 3D-clonogenic survival assay in vitro for radiation alone and for the combination of radiation and temozolomide. GPD1 expression levels were evaluated at protein and mRNA levels. In addition, The Cancer Genome Atlas (TCGA) database was queried for the impact of GPD1 gene expression on survival of GBM patients. R 4.0.3 was used to determine the optimal cutoff value of GPD1. Results Western-blotting results showed that BTSC-tum1 and BTSC-tum6 had higher GPD1 expression than BTSC-A and BTSC-Azer. Based on the results of the growth delay assay and 3D-clonogenic survival assay, BTSC-tum1 and BTSC-tum6 were more resistant to radiation than BTSC-A and BTSC-Azer (p < 0.05). Using the TCGA database, we found that GPD1 was highly expressed in glioblastoma tissues, compared with normal brain tissues. For 501 GBM patients both clinical and GPD1 gene expression data were available in the TCGA database. 275 of these patients underwent radiation therapy with or without chemotherapy. The cutoff value of GPD1 expression was 4.449 log2(affy RMA) (range, 3.973-8.612), 183 tumors showed high GPD1 expression and 92 low expression. Patients with high GPD1 expressing GMB had significantly inferior OS than those with low GPD1 expression (P = 0.012), with median survival times of 15.9 months vs. 20.9 months. Conclusion High GPD1 expression plays a (significant) role in radiation resistance of GBM in preclinical models in vitro and is associated with impaired overall survival of patients receiving radiation therapy with or without chemotherapy. Further preclinical research is currently ongoing to determine the mechanisms by which GPD1 expression confers radiation resistance. PD-0831 Elucidating colony growth in vitro by machine-learning based quantification of time-lapse image data R. Koch 1 , E. Bahn 1,2 , M. Alber 1 1 University Hospital, Department of Radiation Oncology, Heidelberg, Germany; 2 German Cancer Research Centre (DKFZ), Clinical Cooperation Unit Radiation Oncology, Heidelberg, Germany Purpose or Objective In vitro investigation of cellular response to radiation is indispensable in radiobiology. Despite occasional criticism, clonogenicity is defined based on a minimum colony size at a fixed time and is commonly used as a surrogate to quantify the deleterious effect of radiation. Today, automation allows the acquisition and analysis of large time series image data. We used this to augment the classical in vitro clonogenic assay (IVCA) by extracting colony growth (CG) dynamics information on a single-colony level, acquiring time- and dose-dependent colony size distributions from thousands of colonies. Materials and Methods We designed a neural network image segmentation routine based on the U-Net architecture to identify individual colonies in the image data and record morphological parameters. We tracked colonies over time by pairwise matching between subsequent images. We extracted CG parameters using changepoint detection to discern periods of exponential CG and to classify CG behavior utilizing a tree ensemble classifier on a combination of morphological and curve fitting parameters. Data was gathered by seeding cells of different strains (H3122, RENCA, SAT, UT-SCC-5) in multiwell plates or cell culture flasks, X-ray irradiating them at single doses of 0-10 Gy and imaging at regular intervals (3 h or Poster discussions: Poster discussion 18: Tumour-radiobiology