ESTRO meets Asia 2024 - Abstract Book

S321

Physics – Machine learning models and clinical applications

ESTRO meets Asia 2024

metrics and qualitative assessment by an experienced clinical oncologist. These findings suggest the potential of the model to enhance the precision of radiotherapy planning by providing more detailed anatomical information for treatment plan optimization.

Conclusion:

This study contributes to the field of radiotherapy planning by developing a deep learning model that incorporates the segmentation of seminal vesicles (SV) and penile bulb (PB), structures typically absent in publicly available datasets and often excluded in existing segmentation models. By leveraging a pre-trained model on a publicly available dataset and fine-tuning it on a local dataset including SV and PB, this approach demonstrates the feasibility of incorporating these structures for potentially more comprehensive radiotherapy planning. Future work will focus on refining the model's accuracy and generalizability for broader clinical use.

Keywords: Deep learning segmentation, Automatic segmentation

References:

(1) Kanwar, A., Merz, B., Claunch, C., Rana, S., Hung, A., & Thompson, R. F. (2023). Stress-testing pelvic autosegmentation algorithms using anatomical edge cases. Physics and Imaging in Radiation Oncology , 25 , 100413.

(2) Boyd, G. H., Efstathiou, J. A., Kamran, S. C., Zietman, A. L., Miyamoto, D. T., Kirk, M. C., & Wang, Y. (2021). Qualitative and Quantitative Analysis of a Deep Learning Auto Contouring Model for Radiotherapy in Localized Prostate Cancer. International Journal of Radiation Oncology, Biology, Physics , 111 (3), e107-e108.

344

Proffered Paper

Discerning Clinical Target Volume of Endometrial Cancer via a Lightweight Multi- modal Framework

Ang Qu 1 , Lei Zhu 2 , Hang Yang 3 , Weijuan Jiang 1 , Xiuwen Deng 1 , Weiqi Xiong 4 , Ping Jiang 1 , Yanye Lu 5 , Junjie Wang 1

1 Radiation Oncology, Peking University Third Hospital, Beijing, China. 2 College of Future Technology, Peking University, Beijing, China. 3 CRI-BJ-IRTL, United Imaging Research Institute of Intelligent Imaging, Beijing, Beijing, China. 4 RT-IPH, Shanghai United Imaging Healthcare Co., Ltd., Shanghai, China. 5 Institute of Medical Technology, Peking University, Beijing, China

Purpose/Objective:

Postoperative pelvic radiotherapy is a crucial treatment for endometrial cancer. The accurate delineation of the clinical target volume (CTV) represents a critical step in precision radiation therapy. In comparison to non-contrast enhanced computed tomography (NECT), contrast-enhanced computed tomography (CECT) can aid in locating the iliac vessels and contribute to delineating the boundaries of the CTV. Our study aims to assess the impact of CECT on specific delineation of CTV boundaries.

Material/Methods: