ESTRO 2024 - Abstract Book

S2031

Clinical - Paediatric

ESTRO 2024

Fatima Tensaouti 1,2 , Nhi Blanchard 2 , Mathilde Morisseau 3 , Stéphanie Bolle 4 , Alexandre Escande 5 , Xavier Muracciole 6 , Claire Alapetite 7 , Line Claude 8 , Julie Leseur 9 , Jérôme Doyen 10 , Georges Noël 11 , Stéphane Supiot 12 , Valérie Bernier-Chastagner 13 , Julien Welmant 14 , Mathieu Hatt 15 , Soumia Bengoufa 15 , Dominique Figarella Branger 16,17 , Pascale Varlet 18 , Emmanuelle Uro-Coste 19,20,21 , Julien Gautier 22 , Cécile Dalban 22 , Pierre Leblond 23 , Anne Laprie 1,2 1 Oncopole Claudius Regaud, Institut Universitaire du Cancer de Toulouse – Oncopole, Radiation Oncology, Toulouse, France. 2 Toulouse NeuroImaging Center, Université de Toulouse, Inserm, UPS, Research, Toulouse, France. 3 Oncopole Claudius Regaud, Institut Universitaire du Cancer de Toulouse – Oncopole, Biostatistics & Health Data Science Unit, Toulouse, France. 4 Institut Gustave Roussy, Radiation Oncology, Paris, France. 5 Centre Oscar Lambret, Radiation Oncology, Lille, France. 6 La Timone Hospital, Radiation Oncology, Marseille, France. 7 Institut Curie, Radiation Oncology, Paris, France. 8 Centre Léon Bérard, Radiation Oncology, Lyon, France. 9 Centre Eugène Marquis, Radiation Oncology, Rennes, France. 10 Centre Antoine-Lacassagne, University of Côte d’Azur, Radiation Oncology, Nice, France. 11 ICANS - Radiation Oncology, Radiation Oncology, Strasbourg, France. 12 Institut de cancérologie de l’ouest, Radiation Oncology, Nantes, France. 13 Centre Alexis Vautrin, Vandoeuvre, Radiation Oncology, Nancy, France. 14 Institut Cancer de Montpellier, Radiation Oncology, Montpellier, France. 15 LaTIM, Inserm, UMR 1101, Univ Brest, Research, Brest, France. 16 Aix-Marseille University,La Timone Hospital, Pathological Anatomy and Neuropathology, Marseille, France. 17 Institut de Neurophysiopathologie, CNRS - UMR 7051, Research, Marseille, France. 18 GHU Paris-Psychiatrie Et Neurosciences, Sainte-Anne Hospital, Neuropathology, Paris, France. 19 Toulouse University Hospital, Pathology, Toulouse, France. 20 Cancer Research Center of Toulouse (CRCT), INSERM U1037, Research, Toulouse, France. 21 Université Paul Sabatier, Toulouse III, Research, Toulouse, France. 22 Centre Léon Bérard, Clinical Research and Innovation, Lyon, France. 23 IHOP, Lyon, Pediatric Onco Hematology, Lyon, France

Purpose/Objective:

Ependymoma (EPN) is a frequent brain tumor in children. After standard treatment with surgery and radiotherapy, the relapse rate is approximately 40% [1, 2]. No data are available in this type of tumor for risk of relapse stratification using a radiomics approach. The aim of this study was to evaluate the added value of an MRI-derived radiomics signature in stratifying the risk of pediatric EPN in terms of recurrence-free survival (RFS).

Material/Methods:

A total of 345 children aged ≤22 with pathologically confirmed intracranial EPN between 2000 and 2021 and treated with postoperative radiation therapy (RT) were included retrospectively from 13 clinical centers in this study (NCT05151718). MRI examination at diagnosis was available for 211 patients. T1-WI post contrast agent injection was used to manually contour the volumes of interest (VOI) of the tumor. Each T1-WI was preprocessed as follows: 1) voxels interpolation into an isotropic resolution (1×1×1 mm3), 2) discretization of voxels values with a fixed bin number (FBN, 32 bins), and 3) rescaling voxels values within the range of mean value ± (3 × standard deviation inside the VOI). Then, 109 Image Biomarker Standardization Initiative (IBSI)-compliant radiomics features were extracted using the LIFEx 7.3.0 package [3] from each VOI: 42 first-order statistical features; 12 volume and shape features; 55 second-order statistical features. We also included 19 clinical variables and 7 visually derived characteristics [4]. The study endpoint was RFS defined as the time between the initiation of radiotherapy and relapse, death or last follow-up news. Elastic Net penalized Cox proportional hazards regression was used to select the radiomics features associated with RFS. A Rad-score was built with the regression coefficients obtained for the selected