ESTRO 2024 - Abstract Book

S47 ESTRO 2024 Medicine and University Hospital Carl Gustav Carus, Technische Universität Dresden, Helmholtz-Zentrum Dresden, Dresden, Germany. 3 Helmholtz-Zentrum Dresden, Institute of Radiooncology—OncoRay, Dresden, Germany. 4 University of Tübingen, Section for Biomedical Physics, Department of Radiation Oncology, Tübingen, Germany. 5 Icon Cancer Centres, Department of Radiation Oncology, Sydney, Australia. 6 Abano Terme Hospital, Department of Radiation Oncology, Abano, Italy. 7 National Physical Laboratory, Metrology for Medical Physics, Teddington, United Kingdom. 8 Medical University of Vienna, Department of Radiation Oncology, Vienna, Austria. 9 Massachusetts General Hospital, Department of Radiation Oncology, Boston, MA, USA. 10 Paul Scherrer Institute, Center for Proton Therapy, Villigen, Switzerland. 11 Division of Cancer Sciences, The University of Manchester, Manchester, United Kingdom Invited Speaker

Abstract:

Deformable image registration (DIR) is a pivotal tool in modern radiotherapy, facilitating various tasks with its capability to map anatomical changes over a patient's treatment course. This technique finds extensive utility in:

• Image fusion and adaptation: Deforming images for fusion, mid-ventilation images, or online adaptive optimization. • Structure propagation: Facilitating initial contouring from other modalities or atlas segmentation, 4D contouring, or in adaptive radiotherapy workflows. • Dose mapping/accumulation: Enabling the derivation of 4D doses, total dose aggregation from all fractions, or determining doses for treatment combinations and re-irradiation constraints. Despite its versatile promise, DIR encounters persistent challenges due to inherent geometric and subsequent dosimetric uncertainties. Consequently, comprehending the impact and mitigating these uncertainties are paramount for ensuring safe and reliable clinical implementation. This talk will showcase the outcomes of an engaging international ESTRO working group dedicated to discussing DIR uncertainties in radiotherapy. It will delve into the sources of uncertainties, their quantification, and their observed clinical impact, thus shedding light on the current understanding and implications for clinical practice. Unveiling the sources of DIR uncertainties DIR uncertainties stem from a combination of image-based and algorithm-based factors. Image-based uncertainties arise from anatomical changes (e.g., tumor changes, weight fluctuations), artifacts introduced during image acquisition, and the challenges of registering images from different modalities. Algorithm-based uncertainties are linked to the mathematical limitations of DIR algorithms, the choice of similarity metrics, and the influence of user defined parameters. These uncertainties collectively pose potential challenges in various radiotherapy applications. Quantifying DIR uncertainties Accurately quantifying DIR uncertainties is imperative for a comprehensive assessment of their clinical implications, although this task is inherently complex. There is no definitive "ground truth," and different clinical applications (e.g., dose mapping vs. contour propagation) necessitate distinct accuracy needs. Currently, predominantly geometric uncertainty measures have been proposed, with no consensus regarding their suitability for specific applications. Transitioning towards assessing dosimetric impacts, quantification measures remain relatively scarce. Examining the clinical impact The clinical impact of DIR uncertainties is versatile across different applications. A large number of studies reporting geometric DIR uncertainties for structure propagation was summarized, serving as a reference for different anatomical regions and indications. Uncertainties in doses with propagated structures are already much less often reported, as well as dose mapping uncertainties.