ESTRO 2025 - Abstract Book

S3851

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2025

3878

Digital Poster Machine learning integration for prognostic modeling in PSMA-PET-driven salvage radiotherapy for biochemical recurrence post-prostatectomy Federica Medici 1,2 , Daniel M Aebersold 3 , Jozefina Casuscelli 4 , Louise Emmett 5 , Stefano Fanti 6 , Andrea Farolfi 6 , Wolfgang Fendler 7 , Matthias Guckenberger 8 , George Hruby 9 , Stefan A Koerber 10,11 , Stephanie Kroez 8 , Jan C Peeken 12,13 , Paul Rogowski 14 , Sophia Scharl 15 , Mohamed Shelan 16 , Simon K.B. Spohn 17,18,19 , Iosif Strouthos 20 , Lena Unterrainer 21,22 , Marco Vogel 12,23 , Thomas Wiegel 15 , Constantinos Zamboglou 20 , Nina-Sophie Schmidt-Hegemann 14 , Alessio G Morganti 2,24 , Savino Cilla 25 1 Département de Radiothérapie, Gustave Roussy, Paris, France. 2 Department of Medical and Surgical Sciences (DIMEC), Alma Mater Studiorum - Bologna University, Bologna, Italy. 3 Department of Radiation Oncology, Inselspital, Bern University Hospital, Bern, Switzerland. 4 Department of Urology, University Hospital, LMU Munich, Munich, Germany. 5 Department of Theranostics and Nuclear Medicine, St. Vincent's Hospital, Sydney, Australia. 6 Nuclear Medicine, IRCCS Azienda Ospedaliero-Universitaria di Bologna, Bologna, Italy. 7 Department of Nuclear Medicine, University Hospital,University of Essen, Essen, Germany. 8 Department of Radiation Oncology, University Hospital, University of Zurich, Zurich, Switzerland. 9 Department of Radiation Oncology, North Shore Hospital, University of Sydney, Sydney, Australia. 10 Department of Radiation Oncology, Barmherzige Brüder Hospital Regensburg, Regensburg, Germany. 11 Department of Radiation Oncology, Heidelberg University Hospital, Heidelberg, Germany. 12 Deutsches Konsortium für Translationale Krebsforschung (DKTK), Partner Site Munich, Munich, Germany. 13 Department of Radiation Oncology, Klinikum rechts der Isar, Technical University of Munich (TUM), Munich, Germany. 14 Department of Radiation Oncology, University Hospital, LMU Munich, Munich, Italy. 15 Department of Radiation Oncology, University of Ulm, Ulm, Germany. 16 Department of Radiation Oncology, Inselspital, Bern University Hospital, University of Bern, Bern, Switzerland. 17 Department of Radiation Oncology, Medical Center , University of Freiburg, Freiburg, Freiburg, Germany. 18 Deutsches Konsortium für Translationale Krebsforschung (DKTK), Partner Site Freiburg, Freiburg, Germany. 19 Berta-Ottenstein-Programm, Medical Faculty, University of Freiburg, Freiburg, Germany. 20 Department of Radiation Oncology, German Oncology Center, European University Cyprus, Nicosia, Cyprus. 21 Department of Nuclear Medicine, University Hospital, LMU Munich, Munich, Germany. 22 Department of Molecular and Medical Pharmacology, David Geffen School of Medicine, UCLA, Los Angeles, USA. 23 Department of Radiation Oncology, Klinikum rechts der Isar, Technical University of Munich (TUM), Munich, Munich, Germany. 24 Radiation Oncology, IRCCS Azienda Ospedaliero-Universitaria di Bologna, Bologna, Italy. 25 Medical Physics Unit, Responsible Research Hospital, Campobasso, Italy Purpose/Objective: This study aimed to evaluate the prognostic role of PSMA-PET in patients with biochemical recurrence (BCR) following radical prostatectomy (RP) and to assess the impact of salvage radiotherapy (SRT) dosing in patients with or without macroscopic relapse identified by PSMA-PET. The primary endpoints were two-year biochemical relapse free survival (2y-bRFS) and two-year metastasis-free survival (2y-MFS). Material/Methods: Data from 255 patients treated with SRT post-RP were retrospectively collected from 11 medical centers across five countries. Patients with PSA persistence or recurrence (>0.1 ng/mL) were included. Exclusion criteria were pathologically confirmed lymph node metastases, PSMA-PET evidence of nodal or distant metastases, pre-SRT PSA levels >0.5 ng/mL, or prior androgen deprivation therapy (ADT). The primary endpoint was 2y-bRFS. Least Absolute Shrinkage and Selection Operator (LASSO) regression was employed to model the relationship between covariates and clinical outcomes. A Classification and Regression Tree (CART) machine learning model was developed to predict 2y-bRFS. The dataset was split into training/validation (70%) and test (30%) sets using stratified random sampling. Model performance was evaluated using fivefold cross-validation, receiver operating characteristic (ROC) curves, and area under the curve (AUC).