ESTRO 2025 - Abstract Book

S3746

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2025

Conclusion: We developed a multivariable logistic regression model to identify potential risk factors of RICE in paediatric patients treated with PRT. Incorporating age and a location parameter improved model performance. Initial results show that younger patients may be at greater risk for RICE and that this risk may be higher in the brainstem.

Keywords: RICE, NTCP, proton therapy.

References: [1] Bahn, E., Bauer, J., Harrabi, S., Herfarth, K., Debus, J., & Alber, M. (2020). Late Contrast Enhancing Brain Lesions in Proton-Treated Patients With Low-Grade Glioma: Clinical Evidence for Increased Periventricular Sensitivity and Variable RBE. International Journal of Radiation Oncology Biology Physics, 107(3), 571–578. https://doi.org/10.1016/j.ijrobp.2020.03.013 

1436

Digital Poster Temporal validation of [18F]FDG PET-radiomic models for distant-relapse-free-survival after radio chemotherapy for pancreatic adenocarcinoma Monica Maria Vincenzi 1 , Martina Mori 1 , Paolo Passoni 2 , Najla Slim 3 , Emiliano Spezi 4 , Roberta Tummineri 3 , Gabriele Palazzo 1 , Maria Picchio 5,6 , Michele Reni 7,6 , Arturo Chiti 5,6 , Antonella del Vecchio 1 , Claudio Fiorino 1 , Nadia Gisella Di Muzio 3,6 1 Medical Physics, IRCCS San Raffaele Scientific Institute, Milan, Italy. 2 Radiotherapy, IRCCS San Raffaele Scientific Institute, Milano, Italy. 3 Radiotherapy, IRCCS San Raffaele Scientific Institute, Milan, Italy. 4 School of Engineering, Cardiff University, Cardiff, United Kingdom. 5 Nuclear Medicine, IRCCS San Raffaele Scientific Institute, Milan, Italy. 6 Faculty of Medicine and Surgery, Vita-Salute San Raffaele University, Milan, Italy. 7 Oncology, IRCCS San Raffaele Scientific Institute, Milan, Italy Purpose/Objective: Pancreatic cancer ranks among cancers with the poorest prognosis and low survival rates, largely due to frequent diagnoses at advanced (LAPC) or metastatic stages. Aim of current study was to temporally validate a previously developed [18F]FDG-PET radiomic-based model to predict distant relapse-free survival (DRFS). Subsequently, refined models were explored to improve performance. Material/Methods: The original Cox regression model (Model 1) included two radiomic features (RF) and stage (III vs IV). To corroborate and/or refine it, a new study was started aiming to: (i) validate Model 1 in an enlarged cohort of patients; ii) refine Model 1 by checking the robustness of each single RF and iii) evaluate the addition of new, high-performing RFs to develop a fine-tuned, temporally validated index using the same methodology. Data of 215 patients with unresectable LAPC who underwent [18F]FDG-PET/CT and chemoradiotherapy at our Institute between 2005 and 2022 were available. According to internal protocol, stage III and selected stage IV patients in complete response after induction chemotherapy were treated. Patients received concurrent capecitabine (1250 mg/m²/day) and hypo-fractionated IMRT (44.25 Gy/15fr). The cohort was divided into training (145 patients, 2005-2017) and validation (70 patients, 2017-2022). 78 previously identified robust [18F]FDG-PET RFs were extracted using SPAARC-Radiomics Pipeline, according to IBSI guidelines, and then harmonized with the ComBat method. Model 1’s performance was temporally re-tested while the “best” new model was developed using an in-house bootstrap-based machine learning code, similarly to that used in developing Model 1.