S2481
Physics - Radiomics, functional and biological imaging, and outcome prediction
ESTRO 2026
Heumann 3 , David Azria 5 , Ananya Choudhury 6 , Alison M Dunning 7 , Dirk De Ruysscher 8 , Sara Gutiérrez- Enríquez 9 , Ana Vega 10 , Barry S Rosenstein 11 , Elena Sperk 12 , Hilary Stobart 13 , Liv Veldeman 14 , Adam Webb 15 , Christopher J Talbot 15 , Tim Rattay 16 , Jenny Chang-Claude 3 , Catharine West 17 , Tiziana Rancati 1 1 Data Science Unit, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy. 2 Department of Data Science and Biostistics, University Medical Center Utrecht, Utrecht, Netherlands. 3 Division of Cancer Epidemiology, German Cancer Research Center (DKFZ), Heidelberg, Germany. 4 Department of Radiation Oncology, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy. 5 Department of Radiation Oncology, ICM Institut du Cancer Montpellier, Montpelliern, France. 6 Department of Hematology/Oncology, The Christie Hospital NHS Foundation Trust, Manchester, United Kingdom. 7 The Centre for Cancer Genetic Epidemiology, University of Cambridge, Cambridge, United Kingdom. 8 Department of Radiation Oncology, MAASTRO Clinic, Maastricht, Netherlands. 9 Hereditary Cancer Genetics Group, Vall d’Hebron Institute of Oncology, (VHIO), Barcelona, Spain. 10 Department of Cancer Genetics, Fundacion Publica Galega Medicina Xenomica, Santiago de Compostela, Spain. 11 Department of Genetics & Genomic Sciences, Mount Sinai School of Medicine, New York, USA. 12 Department of Radiation Oncology, University Medical Centre, Mannheim, Germany. 13 Patient, Independent Cancer Patients’ Voice (ICPV), UK, United Kingdom. 14 Department of Radiation Oncology, Universiteit Ghent, Ghent, Belgium. 15 Department of Genetics & Cancer Sciences, University of Leicester, Leicester, United Kingdom. 16 Leicester Cancer Research Centre, University of Leicester, Leicester, United Kingdom. 17 Division of Cancer Sciences, University of Manchester, Manchester, United Kingdom Purpose/Objective: Define NT_GARD combining delivered physical dose and individual genetic susceptibility to radiation- induced side-effects. Investigate NT_GARD as a biologically-meaningful exposure-variable in causal inference models, establishing the dose-effect relationship for side-effects after breast cancer (BC) radiotherapy for optimising personalised treatment. Material/Methods: Endpoints: breast fibrosis (CTCAE grade ≥ 2 fibrosis in- tumour-bed/outside-tumour-bed in patients with maximum grade=1 fibrosis pre-radiotherapy).Cohort: 1881 BC patients enrolled in the REQUITE prospective study (with up to 7 years follow-up) with dosimetric, genetic, and toxicity data. Physical dose exposure was defined as the maximum dose to the breast (corrected @2Gy-fractionation, α / β =3Gy, EQD2MaxD) as-per NTCP-model from [Mukesh_2013].To establish patient-
Results: 26 patients were included in the study, equally distributed in gender and with a median age of 60.8 years-old. Proton therapy was delivered in a median of 33 fractions and a prescribed dose of 66 Gy(RBE). After a median follow-up time of 42.3 months, 9 patients reported a local relapse with a 70% (54% - 92%) 3-year local relapse-free survival rate. In our dataset, the measured sphericity (0.5 [0.4, 0.6]) collocates our sample in a medium-high-risk range with respect to Kertels findings. Hence, we focused on the model to discriminate between high and non-high-risk iMM. Among the ten evaluated features, univariable analysis reported a significant impact on the local control only for a texture feature (ClusterProminence, p=0.018). However, 70% of the evaluated radiomics features demonstrated a fair or good 3y-AUROC, with the highest value reported for ClusterProminence (0.85). One first-order (10th percentile), and two shape (flatness and elongation) features reported poor discriminatory capability for local relapse within 3 years. Conclusion: Our preliminary findings reported the potential value of iMM-related radiomics features in predicting high- grade meningioma local relapse within 3 years after proton therapy. However, the evaluation of the signature with potential fine-tuning and larger cohort studies are necessary. References: [1] Maas SLN et al. Integrated Molecular-Morphologic Meningioma Classification: A Multicenter Retrospective Analysis, Retrospectively and Prospectively Validated. JCO 2021;39:3839–52. https://doi.org/10.1200/JCO.21.00784.[2] Kertels O et al. Imaging meningioma biology: Machine learning predicts integrated risk score in WHO grade 2/3 meningioma. Neuro-Oncology Advances 2024;6:vdae080. https://doi.org/10.1093/noajnl/vdae080.[3] Deng MY et al. Analysis of recurrence probability following radiotherapy in patients with CNS WHO grade 2 meningioma using integrated molecular-morphologic classification. Neuro-Oncology Advances 2023;5:vdad059. https://doi.org/10.1093/noajnl/vdad059. Keywords: Radiomics, meningioma, proton-therapy
Mini-Oral 4057
Normal tissue genetically-adjusted radiation dose (NT_GARD) for causal inference models of late toxicity after breast cancer radiotherapy Benedetta Dionisi Ferrera 1 , Eliana Gioscio 1 , Wouter Van Amsterdam 2 , Petra Seibold 3 , Maria Carmen De Santis 4 , Eliana La Rocca 4 , Alessandro Cicchetti 1 , Philipp
Made with FlippingBook - Share PDF online