ESTRO 2025 - Abstract Book

S76

Invited Speaker

ESTRO 2025

and more personalized treatment planning. However, the integration especially in the online workflow is still challenging.

A fully automated MRIgRT workflow with seamlessly integrate AI and advanced imaging technologies, would eliminate the need for manual intervention in many stages and potential streamline the process. In such an integrated system, AI would generate the initial plans, automatically adapt them based on MRI data of the day, perform contouring and synthetic CT generation, and optimize the treatment plan. To achieve this vision, significant collaboration with vendors is necessary. Moreover, seamless integration between MRI systems, treatment planning software, and radiation delivery units is essential for a fully automated workflow. Standardization of workflows, data interoperability, and regulatory considerations must also be addressed to ensure that automated MRIgRT can be safely and efficiently implemented in clinical settings. In conclusion, while AI is already transforming parts of the MRIgRT workflow, a fully automated system remains a work in progress. Advancing AI technologies, improved vendor cooperation, and standardization of processes are key to realizing the full potential of MRIgRT automation, which could significantly enhance treatment precision, reduce clinician workload, and improve patient outcomes

4768

Speaker Abstracts Real-time tracking: The TrackRAD2025 Challenge

Guillaume Landry 1,2,3 , Adrian Thummerer 1 , Christopher Kurz 1 , Coen Hurkmans 4 , Davide Cusumano 5 , Denis Dudáš 6 , Elia Lombardo 1 , Hilary Byrne 7 , Lorenzo Placidi 8 , Marco Fusella 9 , Marco Riboldi 10 , Matteo Maspero 11 , Michael Jameson 7 , Miguel A. Palacios 12 , Paul Keall 13 , Pim Borman 11 , Rob Tijssen 4 , Tom Blöcker 1 , Yiling Wang 14 1 Department of Radiation Oncology, LMU University Hospital, Munich, Germany. 2 German Cancer Consortium (DKTK), partner site Munich, a partnership between DKFZ and LMU University Hospital Munich, Munich, Germany. 3 Bavarian Cancer Research Center, BZKF, Munich, Germany. 4 Department of Radiation Oncology, Catharina Hospital, Eindhoven, Netherlands. 5 Department of Radiation Oncology, Mater Olbia Hospital, Olbia, Italy. 6 Faculty of Nuclear Sciences and Physical Engineering, Czech Technical University in Prague, Prague, Czech Republic. 7 Medical Physics, GenesisCare, Sydney, Australia. 8 Department of Radiation Oncology, Gemelli Hospital, Rome, Italy. 9 Department of Radiation Oncology, Policlinico Abano, Abano Terme, Italy. 10 Department of Medical Physics, LMU Munich, Munich, Germany. 11 Department of Radiation Oncology, UMC, Utrecht, Netherlands. 12 Department of Radiation Oncology, Amsterdam UMC, Amsterdam, Netherlands. 13 ImageX Institute, University of Sydney, Sydney, Australia. 14 Department of Radiation Oncology, Sichuan Cancer Hospital, Chengdu, China The integration of magnetic resonance imaging (MRI) in radiotherapy is becoming increasingly crucial, particularly in patients with tumors susceptible to motion, such as those located in the thoracic, abdominal and pelvic regions. Real-time motion management allows precise targeting of radiation beams, thereby maximizing the dose delivered to the tumor while minimizing exposure to surrounding healthy tissues. One of the most promising advancements in this domain is the development of MRI-guided radiotherapy, specifically utilizing hybrid MRI-linear accelerator (MRI-linac) systems [1], which have been shown to reduce treatment toxicity [2, 3]. These systems enable adaptation to tumor movement during treatment, offering the potential for real-time adjustments in the delivery of radiation. The use of 2D cine-MRI provides a means to visualize tumor motion dynamically, currently allowing the radiation beam to be gated (where radiation delivery is temporarily halted when motion is detected). Ongoing efforts aim at following the tumor with multileaf collimators for improved delivery efficiency while maintaining the same accuracy. Achieving this presents significant challenges due to the need for continuous segmentation of the tumor across all frames of the cine-MRI sequence (also called tumor tracking), typically requiring real-time processing at frame rates Abstract: