ESTRO 2024 - Abstract Book

S5330

Radiobiology - Tumour biology

ESTRO 2024

Cylia Ouadah 1,2 , Soňa Michlíková 1,2 , Alex Zwanenburg 1,3,4 , Artur Yakimovich 5,6 , Nathalie Borgeaud 1,3 , Lydia Koi 1,2 , Safayat Mahmud Khan 7,8 , Maria José Besso 7,8 , Ina Kurth 7,8,9 , Antje Dietrich 1,3 , Mechthild Krause 1,3,10 , Steffen Löck 1,3,10 1 OncoRay - National Center for Radiation Research in Oncology, Faculty of Medicine and University Hospital Carl Gustav Carus, Technische Universität Dresden, Helmholtz-Zentrum Dresden - Rossendorf, Dresden, Germany. 2 Helmholtz-Zentrum Dresden-Rossendorf, Institute of Radiooncology - OncoRay, Dresden, Germany. 3 German Cancer Consortium (DKTK), partner site, Dresden, Germany and German Cancer Research Center (DKFZ), Heidelberg, Germany. 4 National Center for Tumor Diseases (NCT), Partner Site Dresden, Germany: German Cancer Research Center (DKFZ), Heidelberg, Germany, Faculty of Medicine and University Hospital Carl Gustav Carus, Technische Universität Dresden, Dresden, Germany, Helmholtz Association/Helmholtz-Zentrum Dresden-Rossendorf (HZDR), Dresden, Germany. 5 Center for Advanced Systems Understanding (CASUS), Helmholtz-Zentrum Dresden-Rossendorf (HZDR), Görlitz, Germany. 6 Bladder Infection and Immunity Group (BIIG), Department of Renal Medicine, Division of Medicine, University College London, Royal Free Hospital Campus, London, United Kingdom. 7 Division of Radiooncology/Radiobiology, German Cancer Research Center (DKFZ), Heidelberg, Germany. 8 Heidelberg Institute for Radiation Oncology (HIRO) and National Center for Radiation, Research in Oncology (NCRO), Heidelberg, Germany. 9 German Cancer Consortium (DKTK), DKFZ, core center Heidelberg, Heidelberg, Germany. 10 Department of Radiotherapy and Radiation Oncology, Faculty of Medicine and University Hospital Carl Gustav Carus, Technische Universität Dresden, Dresden, Germany The response of locally advanced head and neck squamous cell carcinoma (HNSCC) to (postoperative) radiochemotherapy varies due to tumor heterogeneity. A biomarker-based prediction of tumor radiosensitivity could help to establish personalized treatment strategies and to improve therapy outcome. Histopathological images contain high-dimensional information that can be exploited as source of potential biomarkers utilizing artificial intelligence methods. In particular, the analysis of image data from translationally relevant preclinical animal models of human tumors enables the generation and validation of hypotheses to uncover the biological meaning underlying the biomarker signatures. Therefore, building upon previous experimental findings [1,2], we investigated convolutional neural networks (CNN) for predicting tumor radiosensitivity from histology images of non-irradiated HNSCC xenograft models. Data: In previously published experiments [1,2], athymic nude mice were xenotransplanted with ten HNSCC tumor models of varying radiosensitivity. We classified these models as radioresistant (UT-SCC5, SAS, FaDu, CAL-33 and UT SCC15) and radiosensitive (UT-SCC8, UT-SCC14, SAT, UT-SCC45 and XF354), based on the tumor control dose 50% (TCD50) values from fractionated radiotherapy [1, 2]. Untreated control tumors of approximately 180 mm³ were excised, formalin-fixed and paraffin-embedded. One central slice per tumor specimen was stained with hematoxylin and eosin (H&E) and 71 (4-10 per model) high-resolution whole-slide images (WSI) were obtained. All images were digitized using a 20x magnification and a pixel size of 0.22 µm x 0.22 µm. The dataset was divided into an exploratory and a test set (53 and 18 slides, respectively). Data preparation: Images were first downsampled to 50% of their original resolution, to simplify their processing, while preserving a high quality. Due to the high resolution of the images and the limited sample size, we employed a patch-based method, where each patch is considered as an independent instance with the radiosensitivity classification of the whole slide. Non-overlapping patches of size 1024x1024 were extracted, focusing on regions of Purpose/Objective: Material/Methods: