S238
Clinical - Breast
ESTRO 2026
Institute (CAPHRI), Maastricht, Netherlands. 9 Division of Genetics and Genome Biology, University of Leicester, Leicester, United Kingdom. 10 Leicester Cancer Research Centre, University of Leicester, Leicester, United Kingdom. 11 Department of Radiation Oncology, Gustave Roussy, Villejuif, France. 12 Research, Medical Data Works B.V., Maastricht, Netherlands. 13 Research, Dutch Research Council (NWO), The Hague, Netherlands. 14 Research, Independent Cancer Patients' Voice, London, United Kingdom. 15 Institute of Cognitive Sciences and Technologies (ISTC)Istituto di Scienze e Tecnologie della Cognizione (ISTC), National Research Council (CNR), Rome, Italy
Fatigue trajectories during RT are not uniform but are shaped by treatment history, nodal irradiation, and age. This prospective cohort represents one of the few studies to longitudinally evaluate fatigue and its risk factors in breast cancer patients, providing valuable insight for developing personalized, adaptive fatigue- management strategies in contemporary radiotherapy practice References: 1. Prue G, Rankin J, Allen J, et al.: Cancer-related fatigue: A critical appraisal. Eur J Cancer 42 (7): 846-63, 2006. 2.Mo J, Darke AK, Guthrie KA, et al.: Association of Fatigue and Outcomes in Advanced Cancer: An Analysis of Four SWOG Treatment Trials. JCO Oncol Pract 17 (8): e1246-e1257, 2021. 3. Vaz-Luis I, Di Meglio A, Havas J, et al.: Long-Term Longitudinal Patterns of Patient-Reported Fatigue After Breast Cancer: A Group-Based Trajectory Analysis. J Clin Oncol 40 (19): 2148-2162, 20224.Kang YE, Yoon JH, Park NH, Ahn YC, Lee EJ, Son CG. Prevalence of cancer-related fatigue based on severity: a systematic review and meta- analysis. Sci Rep. 2023 Aug 7;13(1):12815. doi: 10.1038/s41598-023-39046-0. PMID: 37550326; PMCID: PMC10406927 Keywords: Fatigue, Breast cancer, Adjuvant radiotherapy Co-design of a decision-support tool for arm lymphedema risk prediction in breast cancer patients Silvia Gola 1 , Francesca Fracasso 1 , Alessandro Umbrico 1 , Luca Coraci 1 , Aurora Parisi 1 , Iordanis Koutsopoulos 2 , Yannis Thomas 2 , Yiannis Papageorgiou 2 , Alexis Bombezin-Domino 3 , Marie Bergeaud 4 , Guido Bologna 5 , Yuqin Liang 6 , André Panisson 7 , Alan Perotti 7 , Bram Ramaekers 8 , Jordan Rainbird 9 , Tim Rattay 10 , Sofia Rivera 11 , Alessio Romita 12 , Cheryl Roumen 13 , Hilary Stobart 14 , Chris Talbot 9 , Karolien Verhoeven 6 , Gabriella Cortellessa 15 1 Institute of Cognitive Sciences and Technologies (ISTC), National Research Council (CNR), Rome, Italy. 2 Department of Informatics, Athens University of Economics and Business, Athens, Greece. 3 Research, Therapanacea, Paris, France. 4 Research, UNICANCER, Paris, France. 5 Department of Informatics, Hes-so, University of Applied Sciences and Arts of Western Switzerland, Genève, Switzerland. 6 Department of Radiation Oncology (Maastro), GROW-Research Institute for Oncology and Reproduction, Maastricht University Medical Centre+, Maastricht, Netherlands. 7 Research, CENTAI Institute, Turin, Italy. 8 Department of Clinical Epidemiology and Medical Technology Assessment (KEMTA), Maastricht University Medical Centre MUMC+/Care and Public Health Research Digital Poster 1337
Purpose/Objective: As part of the PRE-ACT project (Prediction of
Radiotherapy side Effects using explainable AI for patient Communication and Treatment modification), the aim of our study was to develop an application to communicate to the clinician and the patient the individual AI-predicted risk of arm lymphedema following regional nodal irradiation through a co- design approach involving clinicians and patients. The tool will initially be used as part of a clinical investigation to evaluate its impact on shared decision-making between clinicians and breast cancer patients during the radiotherapy planning phase. Material/Methods: The first phase of the co-design elicited user requirements through 13 interviews with oncologists and focus groups with 34 former breast cancer patients. An initial application mock-up was created based on these requirements and refined through creative workshops with clinicians, covering AI risk representation, recommendations, and informational materials. The application was then developed and integrated with the predictive explainable AI module. The module, including raw model output and graphics of key contributing factors, was evaluated by 30 participants via questionnaires on interpretability and usefulness. Two rounds of usability testing with 13 clinicians led to further design refinements. Results: The application screen shows the individual risk of developing arm lymphedema as a percentage, a linear chart, and an icon array. The two graphics allow comparison between the individual’s risk and those of two reference populations (Figure 1). An explainable AI module visualizes the factors that contribute most to the individual prediction, through a table and a radar chart (Figure 2), and provides personalized recommendations to reduce the risk of developing arm lymphedema. Complementing this, additional guidance on prevention and daily management of arm lymphedema is offered. The AI model’s transparency and explainability is also enhanced by presenting its raw output and details of its training process and
Made with FlippingBook - Share PDF online