S2789
RTT - RTT contouring, target definition, and treatment planning
ESTRO 2026
segmentation, automated treatment planning, and workflow optimization Roel Kierkels, Joop Woltman, Lindsy Adema, Marloes Spoolder - ten Brinke, Paul Jeene Department of Radiation Oncology, Radiotherapiegroep, Arnhem/Deventer, Netherlands
Results: The table shows the organ-specific and the overall accuracy of the models trained and tested on different datasets. The T1-specific model showed significantly (P<0.005) better accuracy (87.57%) on T1(C) data than the hybrid one (86.70%), while it had significantly (P<0.005) lower accuracy (82.44%) on T1 data compared to the hybrid model (83.48%). Similarly, the hybrid model showed significantly (P<0.02) better accuracy (83.29%) on T2 data compared to the T2- specific one (82.75%). The better overall accuracy measured on T1(C) data was due to the higher resolution, while the better accuracy of the hybrid model on T1 and T2 data was due to the diverse training data.
Conclusion: Experiments with a state-of-the-art segmentation model demonstrated it is beneficial to use various MR sequences to train generalized organ segmentation models, which can even out-perform the sequence- specific models.
Digital Poster 3247
Dosimetric Impact of Deep Inspiration Breath Hold in Breast, Axillary, and Internal Mammary Chain Radiotherapy Daniela Branco 1 , Fernando M. Costa 1,2 , Raquel Rocha 1 , Daniela Saraiva 1,3 , Armanda Monteiro 1,2 , Lígia Osório 1 1 Radiotherapy, ULS São João, Porto, Portugal. 2 Medical Imaging and Radiotherapy, E2S - P. Porto, Porto, Portugal. 3 University of Vigo, University of Vigo, Vigo, Spain Purpose/Objective: Deep Inspiration Breath Hold (DIBH) is a well- established technique in breast cancer radiotherapy, aiming to minimise radiation exposure to the heart and lungs, particularly for left-sided treatments. However, its dosimetric benefits warrant further quantification when irradiation includes the internal mammary chain (IMC) and supraclavicular nodes.The aim of this study is to assess the impact of DIBH on cardiac and pulmonary doses in patients undergoing
References: [1] Isensee, F., Jaeger, P.F., Kohl, S.A.A. et al. nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nat Methods 18, 203–211 (2021). https://doi.org/10.1038/s41592-020- 01008-z Keywords: deep-learning, segmentation, magnetic- resonance Proffered Paper 3229 Rapid palliative radiotherapy with deep learning
Made with FlippingBook - Share PDF online