ESTRO 2025 - Abstract Book

S2439

Physics - Autosegmentation

ESTRO 2025

Conclusion: The trainable morphological post-processing framework effectively improves thoracic OAR segmentation accuracy. Its seamless integration with various deep learning models and consistent performance enhancement highlight its potential for clinical adoption. Future work will explore advanced morphological operations and extend the framework to additional segmentation tasks to further validate its adaptability and utility.

Keywords: autosegmentation, organ at risk

References: Kidd, A. C., Skrzypski, M., Jamal-Hanjani, M., & Blyth, K. G. (2019). Cancer cachexia in thoracic malignancy: A narrative review. Current Opinion in Supportive and Palliative Care, 13(4), 316-322. Lambert, Z., Petitjean, C., Dubray, B., & Kuan, S. (2020, November). Segthor: Segmentation of thoracic organs at risk in ct images. In 2020 Tenth International Conference on Image Processing Theory, Tools and Applications (IPTA) (pp. 1-6). IEEE. Franchi, G., Fehri, A., & Yao, A. (2020). Deep morphological networks. Pattern Recognition, 102, 107246.

1621

Mini-Oral Assessment of a novel approach for validating auto-contouring systems with an expert predicted acceptability range Katherine Mackay 1,2,3 , David Bernstein 4,2 , Sooha Kim 2 , Ben Glocker 5 , Binnaz Yasar 1,2 , Leslie Cheng 1,2 , Stephen D Robinson 6,7 , Vincent Khoo 1,2 , Alexandra Taylor 1,2 1 Department of Radiotherapy, The Royal Marsden Hospital, London, United Kingdom. 2 Radiotherapy and Imaging, The Institute of Cancer Research, London, United Kingdom. 3 NHS Clinical Fellowship in AI, NHS, London, United Kingdom. 4 Department of Physics, The Royal Marsden Hospital, London, United Kingdom. 5 Department of Computing, Imperial College London, London, United Kingdom. 6 Sussex Cancer Centre, University Hospitals Sussex NHS Foundation Trust, Brighton, United Kingdom. 7 Department of Biochemistry, University of Sussex, Brighton, United Kingdom