ESTRO 2025 - Abstract Book

S3744

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2025

Keywords: Classification, Regression

References: 1. Sharabiani M et al. Generalizability assessment of head and neck cancer NTCP models based on the TRIPOD criteria. Radiother Oncol. 2020 May 1;146:143–50. 2. Van den Bosch L et al. Comprehensive toxicity risk profiling in radiation therapy for head and neck cancer: A new concept for individually optimised treatment. Radiother Oncol J Eur Soc Ther Radiol Oncol. 2021 Apr;157:147–54. 3. Nielsen CP et al. Consistency in contouring of organs at risk by artificial intelligence vs oncologists in head and neck cancer patients. Acta Oncol Stockh Swed. 2023 Nov;62(11):1418–25. 4. Vergouwe Y et al. A closed testing procedure to select an appropriate method for updating prediction models. Stat Med. 2017;36(28):4529–39.

1429

Poster Discussion Development of a prediction model for radiation-induced contrast enhancement after proton therapy for brain tumours in paediatric patients Abel Bregman 1 , Jikke J. Rutgers 2 , Truls Andersen 1 , Arjen van der Schaaf 1 , John H. Maduro 1 , Charlotte L. Brouwer 1 , Geert O. Janssens 3,4 , Eelco W. Hoving 2 , Hiske L. van der Weide 1 , Maarten H. Lequin 5 , Rutger A.J. Nievelstein 5,6,7 , Stefan Both 1 , Johannes A. Langendijk 1 , Dirk Wagenaar 1 1 Department of Radiation Oncology, University of Groningen, University Medical Center Groningen, Groningen, Netherlands. 2 Department of Neuro-Oncology, Princess Máxima Center for Paediatric Oncology, Utrecht, Netherlands. 3 Department of Radiation Oncology, University Medical Center Utrecht, Utrecht, Netherlands. 4 Department of Paediatric Oncology, Princess Máxima Center for Paediatric Oncology, Utrecht, Netherlands. 5 Department of Radiology and Nuclear Medicine, University Medical Center Utrecht, Utrecht, Netherlands. 6 Department of Radiology and Nuclear Medicine, Wilhelmina Children's Hospital, Utrecht, Netherlands. 7 Department of Radiology and Nuclear Medicine, Princess Máxima Center for Paediatric Oncology, Utrecht, Netherlands Purpose/Objective: The occurrence of radiation-induced contrast enhancement (RICE) after brain radiotherapy has been identified as a precursor to clinically manifested toxicity. This study aimed to develop a multivariable prediction model for RICE after proton therapy (PRT) in paediatric patients with posterior fossa tumours, and to compare this model with a previously published model for adults. Material/Methods: This retrospective study included 67 paediatric patients. Prescribed doses were 54.0 Gy RBE(=1.1) or 59.4 Gy RBE(=1.1) . Each patient had three-monthly follow-up MRI scans evaluated by a paediatric radiologist for RICE. RICE was delineated at its first occurrence. Delineations were propagated to the planning CT (pCT). Dose, dose•LETd, OAR and RICE data of pCT voxels that received ≥2.0 Gy RBE(=1.1) were extracted. A previously published model for predicting RICE at the voxel level in adult patients[1], incorporating dose, dose•LETd and a parameter for the proximity to the ventricular system, was evaluated in our cohort and refitted. Thereafter, we developed a multivariable logistic regression model incorporating parameters for the dose per voxel, age and the location of the brainstem (Figure 1). Model development was performed using forward selection and the optimism corrected likelihood. The models were used to predict RICE at the voxel level, such that information from voxels receiving ≥2.0 Gy RBE(=1.1) was included. We calculated the probability of lesion origin (POLO). NTCP was obtained by combining the POLO of included voxels and thereafter used to assess model performance by calculating the AUC and Brier score. Bootstrapping was implemented to obtain 95%CI values for parameters and performance metrics.