ESTRO 2025 - Abstract Book

S2528

Physics - Autosegmentation

ESTRO 2025

Conclusion: The nnUNet model for RC segmentation demonstrated robust performance, achieving a DSC score comparable to the range of interrater variability. Despite the small training cohort of 24 patients, the dura segmentation model delivered promising results. Expanding the training datasets for both models is expected to enhance their performance. However, quantitative measures for the automatic segmentation of the CTV and its clinical usefulness still must be assessed.

Keywords: U-Net, resection cavity, CTV

References: [1] Soliman H, et al. Consensus Contouring Guidelines for Postoperative Completely Resected Cavity Stereotactic Radiosurgery for Brain Metastases. Int J Radiat Oncol Biol Phys. 2018; 100(2):436-442. [2] Ronneberger, Olaf, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part III 18. Springer International Publishing, 2015.

3494

Digital Poster Implementation of User-Centric Dashboard to Automatically Monitor the Dynamics of Lung Cancer Cachexia using AI-driven analysis of 3D CBCT images Behzad Rezaeifar 1 , Wouter R P H van de Worp 2 , Sara I S Manso 1 , Lars H B A Daenen 1 , Cecile J A Wolfs 1 , Peiyu Qiu 2 , Joël de Bruijn 3 , Justine M Webster 2 , Stéphanie Peeters 1 , Dirk De Ruysscher 1 , Ramon C J Langen 2 , Frank Verhaegen 1,3 1 Radiation Oncology (Maastro), GROW Research Institute for Oncology and Reproduction, Maastricht, Netherlands. 2 Respiratory Medicine, NUTRIM Institute of Nutrition and Translational Research in Metabolism, Maastricht, Netherlands. 3 SmART Scientific Solutions BV, -, Maastricht, Netherlands Purpose/Objective: Cachexia affects 39% of patients with locally advanced non-small cell lung cancer (NSCLC)[1]. It significantly impairs the tolerance for treatment and survival. Currently, cachexia is clinically assessed through body weight measurements, which do not provide in-depth insights into the tissue wasting process. Furthermore, early identification of cachexia during treatment bears the potential for personalized interventions. Therefore, we monitored the dynamics of cachexia by leveraging AI models to automatically contour and analyze various organs, different muscles, and adipose tissues during the course of treatment, using 3D cone-beam CT (CBCT) images at each radiotherapy fraction.