ESTRO 2025 - Abstract Book

S3722

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2025

530

Digital Poster SBRT in oligometastatic patients: radiomics for predicting local control

Carolina De la Pinta 1,2 , Laura Romero 3 , Carmen Vallejo 4 , Teresa Muñoz 4 , Sonsoles Sancho 5 , Maria Elena Hernando 6 1 Radiation Oncology Department, Ramón y Cajal Hospital, Madrid, Spain. 2 Biomarkers and Therapeutic Targets Group, IRYCIS, Madrid, Spain. 3 Radiation Oncology Department, IRYCIS, Madrid, Spain. 4 Radiation Oncology Department, Ramon y Cajal Hospital, Madrid, Spain. 5 Radiation Oncology, Ramón y Cajal Hospital Hospital, Madrid, Spain. 6 Bioengineering and Telemedicine Group, Polytechnic University of Madrid, Madrid, Spain Purpose/Objective: To explore the potential of computed tomography (CT) in predicting treatment response to Stereotactic Body Radiation Therapy (SBRT) for patients with oligometastases from breast and prostate cancer. Material/Methods: A total of 114 metastatic lesions from 78 patients treated between 2016 and 2024 were analyzed in this study. Response to SBRT divided patients with Complete Response (CR) and Partial Response (PR); and Stable Disease (SD) and Progressive Disease (PD) according to RECIST 1.1 criteria. A total of 160 radiomic features (RFs) were extracted from CT images before the treatment. RFs were selected using Mann-Whitney U/Wilcoxon rank sum test for continuous non-normally distributed variables and Welch t test for normally distributed and non-homoscedastic variables. Highly correlated features were removed using Spearman's correlation test. Six machine learning algorithms were trained to predict response to SBRT (CR and PR), using stratified 5-fold cross validation and Recursive Feature Elimination (RFE). Results: Patients' median age was 67.5 years (36-95); 49 patients were male. Median total dose was 30Gy(16-45Gy) in 3 fractions (1-6). Response to SBRT was distributed as 33-9-66-6 lesions (CR-PR-SD-PD). Median Overall Survival (OS) was 29.82 months (1.97-82.63 months). Spearman's correlation test was applied to 147 clinical and radiomic features that were statistically different between groups. Four features were selected with the RFE to train the models. The best performing model after harmonization by the Spiral Pitch Factor Tag (0018,9311) with COMBAT was the SVM achieving a Receiver Operating Curve AUC of 0.8; a precision of 0.77 and a recall of 0.71. Based on these findings, a predictive nomogram has been created based on clinical and radiomic characteristics. Conclusion: CT-based radiomics has the potential to provide reasonable biomarkers of response to SBRT treatment for metastases, and it can help improve clinical decisions for early treatment adaptation. A nomogram of response prediction has been created.

Keywords: RADIOMICS, SBRT, oligometastasic

686

Digital Poster Hypoxia mapping from diffusion MRI to differentiate patients with recurrence after peripheral zone prostate cancer radiotherapy Valentin Septiers 1,2 , Maria A Zuluaga 2 , Carlos S Marrero 1 , Aurélien Briens 1 , Léo Le Bozec 1 , Anaïs Barateau 1 , Hervé Saint-Jalmes 1 , Renaud De Crevoisier 1 , Oscar Acosta 1 1 LTSI - UMR 1099, Univ Rennes - CLCC Eugène Marquis - INSERM, Rennes, France. 2 Data Science Department, EURECOM, Biot, France