ESTRO 2025 - Abstract Book

S3380

Physics - Machine learning models and clinical applications

ESTRO 2025

Conclusion: The target volumes auto-segmentation models were tailored to align with the rectal cancer target volumes segmental conventions established by Peking University Third Hospital. The resultant target volume segmentation adhered to the segmental standards of our center, thereby providing preliminary validation of the clinical applicability of the three auto-segmentation models for target volumes. The methodology employed for model training holds potential for adaptation across other medical centers, which may enhance its generalizability. Furthermore, the CT-based multiple auto-segmentation models demonstrate the capacity to enhance the efficiency, consistency, and accuracy of target volume segmentation in rectal cancer radiotherapy, thereby indicating a superior clinical application value. References: [1] Yu L, Zhao J, Zhang Z, Wang J, Hu W. Commissioning of and preliminary experience with a new fully integrated computed tomography linac. J Appl Clin Med Phys. 2021;22(7):208-223 [2] Yap MH, Cassidy B, Byra M, et al. Diabetic foot ulcers segmentation challenge report: Benchmark and analysis. Med Image Anal. 2024;94:103153 [3] E. Giacomello, D. Loiacono and L. Mainardi. Brain MRI Tumor Segmentation with Adversarial Networks. 2020 International Joint Conference on Neural Networks (IJCNN), Glasgow, UK. 2020;1-8 [4] N. Ibtehaz, M.S. Rahman, MultiResUNet: Rethinking the U-net architecture for multimodal biomedical image segmentation, Neural Netw. 2020;121;74 – 87 Keywords: Auto-segmentation; Deep learning; Rectal cancer

1456

Digital Poster Clinical priorities-driven head and neck simultaneous integrated boost radiotherapy using deep reinforcement learning Dongrong Yang, Xin wu, xinyi li, yibo xie, qiuwen wu, yang sheng, Q. Jackie wu radiation oncology, Duke University Medical Center, Durham, USA Purpose/Objective: Head and neck (HN) simultaneous-integrated-boost(SIB) radiotherapy planning is particularly challenging due to proximate organs-at-risk (OARs)(1). Depending on the specific clinical conditions, distinctive parotid sparing strategies are utilized to preserve parotids function without compromising target coverage. We propose a deep-