ESTRO 2020 Abstract Book

S859 ESTRO 2020

corresponds to 50% of the total dose at each voxel. For each patient, BED distributions were calculated in a region covering the irradiated target based on the linear- quadratic model with and without taking into account the effects of dose-rate, the time between beams and the overall treatment time. Slow and fast components of repair were incorporated into the BED formulation that takes into account the aforementioned time-depended terms. Results The target average second quartile dose-rates of the studied cases were found equal to 2.49 ± 0.34 and 3.11 ± 0.43 Gy/min for the 800 and 1000 MU/min output values, respectively. Averaged BED values of 97 ± 3 Gy 2.47 covering 95% of the target volume were found without taking account of slow and fast repairs, while dropped down to 78 ± 3 Gy 2.47 for the 800 MU/min nominal dose-rate when time-dependent parameters were considered. A dependence of 2.5% was found between the 800 and 1000 MU/min linac output rates. Conclusion A BED loss of approximately 20% was found in robotic radiosurgery treatments when dose rate effects were considered. This BED loss was found to improve by 2.5% when the 800 MU/min linac was replaced with the 1000 MU/min one. PO-1583 Non-invasive radiomic imaging prediction of tumour hypoxia: biomarker for FLASH irradiation? S. Sanduleanu 1 , T. Tamanupadhaya@gmail.com 2 , R. Klaassen 3 , H. Woodruff 4 , M. Hatt 5 , J. Kaanders 6 , O. Vrieze 7 , H. Laarhoven 8 , R. Subramiam 9 , S.H. Huang 10 , S. Bratman 11 , L. Dubois 12 , R. Miclea 13 , D. Di Perri 14 , X. Geets 15 , M. Crispin-Ortuzar 16 , A. Aptea 17 , J. Hun Oh 18 , N. Lee 18 , J. Humm 18 , H. Schoder 18 , D. Ruysscher 19 , F. Hoebers 19 , P. Lambin 1 1 Maastricht University, Precision Medicine, Maastricht, The Netherlands ; 2 University of California San Francisco UCSF Medical Center, Radiation Oncology, San Francisco, USA ; 3 AMC Medical Center, Medical Oncology, Amsterdam, Belgium ; 4 Maastricht University, Precision Medicine, Maastricht, Belgium ; 5 French Institute of Health and Medical Research, Unit of Continuous Optimization of Therapeutic Actions through the Integration of Multimodal Information, Brest, France ; 6 Radboud hospital, Radiation Oncology, Nijmegen, Belgium ; 7 Antoni van Leeuwenhoek Medical Centre, Radiation Oncology, Amsterdam, Belgium ; 8 AMC Academic Medical Centre, Medical Oncology, Amsterdam, Belgium ; 9 University of Texas Southwestern Medical Centre, Department of Radiology and Radiologic Sciences, Dallas, USA ; 10 Princess Margaret Cancer Center, Radiation Oncology, Toronto, Canada ; 11 Princess Margaret Cancer Centre, Radiation Oncology, Toronto, Canada ; 12 Maastricht University, Department of Precision Medicine- The M-LAB, Maastricht, Belgium ; 13 Maastricht University Medical Centre+, Radiology, Maastricht, Belgium ; 14 Cliniques universitaires Saint- Luc, Radiation Oncology, Ziekenhuis in Sint-Lambrechts- Woluwe, Belgium ; 15 Cliniques Universitaires St-Luc, Radiation Oncology, Brussels, Belgium ; 16 University of Cambridge, Cancer Research UK Cambridge, Cambridge, United Kingdom ; 17 Memorial Sloan Kettering Cancer Center, The Joseph Deasy Lab, New York, USA ; 18 Memorial Sloan Kettering Cancer Center, Radiation Oncology, New York, USA ; 19 Maastro Clinic, Radiation Oncology, Maastricht, Belgium Purpose or Objective Tumor hypoxia increases resistance to radiotherapy and systemic therapy. The FLASH effect, demonstrating a remarkable sparing of normal tissue after irradiation at ultra-high dose rate (>40Gy s -1 ), is hypothesized to deplete oxygen too quickly for diffusion to maintain an adequate

level of oxygenation, and consequently the normal tissue will respond as a hypoxic tissue. When a hypoxic (radioresistant) tumour is surrounded by an oxic tissue (radiosensitive), ultra-high dose rate will increase the radioresistance of the normal tissue with small impact on the already hypoxic tumour tissue. Our aim was to develop and validate an agnostic and site-specific CT and FDG-PET- based radiomics hypoxia classification signature as a biomarker for future (pre)-clinical FLASH trials. Material and Methods A total of 808 patients from 8 registered prospective clinical trials were included in the generation and validation of the CT and FDG-PET hypoxia classification signature. The primary gross tumor volumes (GTV) were manually defined on CT. In order to dichotomize between hypoxic/well-oxygenated tumors threshold values of 10%, 20% and 30% were used for the [ 18 F]-HX4-derived hypoxic fractions (HF, defined as the ratio between the high uptake hypoxic regions with tumor-to-background ratio>1.4 to total GTV). A random forest (RF)-based machine-learning classifier was trained to classify patients as hypoxia-positive/ negative based on radiomic features. Results An agnostic CT model reached AUC’s of 0.79±0.16, 0.76±0.18 and 0.72±0.14 respectively in external validation by combining 5 CT-derived radiomic features to classify hypoxia (HF cutoff 20%). An agnostic FDG-PET model reached an AUC of 0.74±0.23 in external validation by combining 10 features (HF cutoff 20%). The lung- specific model reached an AUC of 0.80±0.15 in external validation with 4 CT features, while the H&N-specific model reached an AUC of 0.86±0.20 in external validation with 6 CT features. A significant survival split (P=0.037) was found between CT-classified hypoxia strata in an external H&N cohort (n=517), while 117 significant hypoxia gene-CT signature feature associations were found in an external lung cohort (n=80). Conclusion Our hypoxia signatures have the potential to enrich clinical FLASH trials by identifying patients with tumors likely to be hypoxic. PO-1584 Predicting xerostomia in head and neck cancer using imaging biomarkers from daily tomotherapy MVCTs T. Berger 1 , D.J. Noble 2 , L.E. Shelley 1 , T. McMullan 1 , M. Romanchikova 3 , L.J. Carruthers 1 , L. Cebamanos 4 , G. Beckett 4 , A. Duffton 5 , C. Paterson 5 , R. Jena 2 , D.B. McLaren 6 , N.G. Burnet 7 , W.H. Nailon 8 1 Edinburgh Cancer Centre, Department of Oncology Physics, Edinburgh, United Kingdom ; 2 The University of Cambridge, Department of Oncology, Cambridge, United Kingdom ; 3 National Physical Laboratory, Data Science, Teddington, United Kingdom ; 4 Edinburgh Parallel Computing Centre, High Performance Computing, Edinburgh, United Kingdom ; 5 Beatson West of Scotland Cancer Centre, Oncology, Glasgow, United Kingdom ; 6 Edinburgh Cancer Centre, Department of Clinical Oncology, Edinburgh, United Kingdom ; 7 Division of Cancer Sciences- University of Manchester, Manchester Academic Health Science Centre- and The Christie NHS Foundation Trust, Manchester, United Kingdom ; 8 Edinburgh Cancer Centre, Department of Oncology Physics / School of Engineering- the University of Edinburgh, Edinburgh, United Kingdom Purpose or Objective There is a growing interest in predicting late xerostomia in head and neck (H&N) cancer patients using image biomarkers (IMBs) calculated on CT images acquired at sparse intervals over the course of radiotherapy. In delivering radiotherapy with image guidance (IG) the TomoTherapy (Accuray, Sunnyvale, CA, USA) system allows daily IG mega-voltage CT (MVCT) scans to be acquired with ease. The aim of this study was to 1) investigate whether