ESTRO 2023 - Abstract Book

S1890

Digital Posters

ESTRO 2023

PO-2106 Machine learning prediction of pain response to palliative radiation therapy with CT-based radiomics

O. Llorian 1,2,3 , J. Akhgar 1 , S. Pigorsch 1 , K. Borm 4 , S. Münch 4 , D. Bernhardt 4,5,6 , B. Rost 7 , M. Andrade 8 , S. Combs 4,9,6 , J. Peeken 4,9,6 1 Klinikum rechts der Isar, Technical University of Munich, Department of Radiation Oncology, Munich, Germany; 2 School of Computation, Information and Technology, Technical University of Munich, Department for Bioinformatics and Computational Biology, Garching, Germany; 3 Institute of Organismic and Molecular Evolution, Johannes Gutenberg University Mainz , Computational Biology and Data Mining, Mainz, Germany; 4 Klinikum rechts der Isar, Technical University of Munich , Department of Radiation Oncology, Munich, Germany; 5 Institute of Radiation Medicine, Helmholtz Zentrum, Department of Radiation Sciences, Munich, Germany; 6 Deutsches Konsortium für Translationale Krebsforschung, Deutsches Konsortium für Translationale Krebsforschung, Munich, Germany; 7 School of Computation, Information and Technology, Technical University of Munich , Department for Bioinformatics and Computational Biology, Garching, Germany; 8 Institute of Organismic and Molecular Evolution, Johannes Gutenberg University Mainz, Computational Biology and Data Mining, Mainz, Germany; 9 Institute of Radiation Medicine, Helmholtz Zentrum , Department of Radiation Sciences, Munich, Germany Purpose or Objective Painful Spinal Bone Metastases (PSBMs) patients regularly receive palliative Radiation Therapy (RT) with response rates in about 2/3 of patients. In this study, we evaluated the value of Machine Learning (ML) models based on radiomics and semantic imaging features, as well as clinical parameters to predict complete pain response. Materials and Methods Gross Tumour Volumes (GTV) and Clinical Target Volumes (CTV) of 261 PSBMs were segmented on planning computed tomography (CT) scans. 105 radiomics features were extracted and pre-processed from both volumes of interest. Semantic features from the Spinal Instability Neoplastic Score (SINS), and clinical features, were determined and collected for all patients. ML techniques, including random forest classifier (RFC) and support vector machine (SVM), were trained on the radiomics, semantic and clinical features, and compared using repeated nested cross validation. Models trained on combined features were also evaluated for a possible performance increase. Feature importance was analysed via the mean decrease in impurity for RFC models. Results The best radiomics classifier was trained on CTV with an Area Under the Receiver-Operator Curve (AUROC) of 0.62 ± 0.01 (RFC). The semantic model achieved a comparable AUROC of 0.63 ± 0.01 (RFC). This was significantly lower than the clinical ML model (SVM, AUROC: 0.80 ± 0.01) and slightly lower than the Spinal Instability Neoplastic Score (SINS; LR, AUROC: 0.65 ± 0.01). A combined SVM model trained on CTV, SINS, and clinical features did not further improve performance (AUROC: 0,74 ± 0,01). Feature selection frequency and importance analysed on combined models, confirmed that clinical and semantic features were selected more often, achieving high importance scores. Conclusion In this exploratory study, we could demonstrate that radiomics and semantic analyses of planning CTs allowed for prediction of therapy response to palliative RT, albeit to a limited extent. The best prediction was possible by applying ML modelling to established clinical parameters, which also improved the performance of the combined models that used them. L. Humbert-Vidan 1,2 , E. Blackmore 3 , C.R. Hansen 4 , C.D. Fuller 5 , S. Petit 6 , A. van der Schaaf 7 , L.V. van Dijk 7 , G.M. Verduijn 6 , H. Langendijk 7 , C. Muñoz-Montplet 8,9 , W. Heemsbergen 6 , M. Witjes 10 , A.S.R. Mohamed 5 , A.A. Khan 11 , J. Marruecos Querol 9,12 , I. Oliveras Cancio 12 , V. Patel 13 , A.P. King 14 , J. Johansen 15 , T. Guerrero Urbano 16 1 Guy's and St Thomas' NHS Foundation Trust, Radiotherapy Physics, London, United Kingdom; 2 King's College London, Comprehensive Cancer Centre, School of Cancer and Pharmaceutical Sciences, London, United Kingdom; 3 King's College London, Research Management & Innovation Directorate, London, United Kingdom; 4 University of Southern Denmark, Department of Clinical Research, Odense, Denmark; 5 The University of Texas MD Anderson Cancer Centre, Department of Radiation Oncology, Houston, Texas, USA; 6 Erasmus MC Cancer Institute, University Medical Centre Rotterdam, Department of Radiotherapy, Rotterdam, The Netherlands; 7 University Medical Centre Groningen, University of Groningen, Department of Radiation Oncology, Groningen, The Netherlands; 8 Catalan Institute of Oncology, Department of Medical Physics and Radiation Protection, Girona, Spain; 9 University of Girona, Department of Medical Sciences, Girona, Spain; 10 University Medical Centre Groningen, University of Groningen, Department of Radiation Oncology, , Groningen, The Netherlands; 11 Odense University Hospital, Department of Oral and Maxillofacial Surgery, Odense, Denmark; 12 Catalan Institute of Oncology, Department of Radiation Oncology, Girona, Spain; 13 Guy’s and St Thomas’ NHS Foundation Trust, Department of Oral Surgery, London, United Kingdom; 14 King’s College London, School of Biomedical Engineering and Imaging Sciences, London, United Kingdom; 15 Odense University Hospital, Department of Oncology, Odense, Denmark; 16 Guy’s and St Thomas’ NHS Foundation Trust, Department of Clinical Oncology, London, United Kingdom Purpose or Objective Multi-centre real-world data research collaborations are essential in radiation-induced toxicity prediction modelling in order to achieve larger and more diverse datasets, especially for rare toxicities such as mandibular osteoradionecrosis (ORN). The PREDMORN international multi-centre study was developed to achieve more statistically robust and generalisable conclusions than the existing published studies on ORN modelling. We describe the main challenges and limitations encountered in the data sharing process and make some recommendations based on the PREDMORN study experience. PO-2107 Challenges in international real world evidence research collaboration. The PREDMORN experience