S2107
Physics - Image acquisition and processing
ESTRO 2026
in South China, Guangdong Key Laboratory of Nasopharyngeal Carcinoma Diagnosis and Therapy, Guangdong Provincial Clinical Research Center for Cancer, Sun Yat-sen University Cancer Center, Xiamen, China. 3 Department of Radiation Oncology, The Third Affiliated Hospital of Sun Yat-sen University, Guangzhou, China. 4 Department of Radiation Oncology, the First People’s Hospital of Foshan, Foshan, China Purpose/Objective: In thoracic radiotherapy, respiratory motion induces positional shifts in tumor targets, resulting in discrepancies between planned and delivered dose distributions. These deviations may reduce therapeutic efficacy and elevate complication risk. While X-ray-based image guidance is effective, daily application is constrained by cumulative radiation exposure. Given the correlation between external surface motion and internal respiratory movement, establishing a method to accurately predict the spatial position of targets dynamic using surface motion characteristics can provide a technical solution for precise irradiation of intrathoracic moving tumors guided by optical surface imaging. This study aims to develop a deep learning model based on the 3D nnU- Net architecture to achieve real-time prediction and monitoring of gross tumor volume (GTV) deformation in lung cancer patients by correlating surface motion with internal tumor dynamics. Material/Methods: The study included 119 4DCT datasets from patients with lung cancer, divided into a training set (n=75), an internal test set (n=16), and two external test sets (n=13 and n=15). External/internal motion data were extracted from GTV/body contours on 4DCT images. A 3D nnU-Net-based model was trained to map external surface motion to internal GTV deformation. To optimize model efficiency and accuracy, the impact of different surface input regions was systematically evaluated to identify the optimal surface area. Model performance was measured using Dice similarity coefficient, 95th percentile Hausdorff distance (HD95), and mean surface distance (MSD). Results: The model accurately predicted three-dimensional GTV contours in lung cancer. Internal test results showed a mean Dice of 0.89±0.11, MSD of 0.92±0.86 mm, and HD95 of 3.17±3.03 mm. The model also maintained excellent generalization performance on two independent external test sets, achieving Dice coefficients of 0.88±0.12 and 0.90±0.11. Furthermore, the model's accuracy improved significantly when the input surface region radius was expanded from 50 mm to 100 mm (P<0.05), matching full-surface input accuracy (P>0.05). Conclusion:
yielded comparable values within ~1%. The head phantom used for evaluation does not have an oral cavity.
Conclusion: The proposed Sim2Real approach enables effective, physically grounded scatter correction in CBCT, outperforming raw and purely simulated methods. By integrating experimental calibration data for domain adaptation, the model ensures improved image fidelity and auto-contouring accuracy, advancing the clinical readiness of deep learning–based CBCT correction for adaptive radiotherapy. References: [1] Lalonde A, Winey B, Verburg J, Paganetti H, Sharp GC. Evaluation of CBCT scatter correction using deep convolutional neural networks for head and neck adaptive proton therapy. Phys Med Biol. 2020 Dec 4;65(24):10.1088/1361-6560/ab9fcb.[2] Niu T, Sun M, Star-Lack J, Gao H, Fan Q, Zhu L. Shading correction for on-board cone-beam CT in radiation therapy using planning MDCT images. Medical Physics. 2010;37(10):5395–406.[3] Park YK, Sharp GC, Phillips J, Winey BA. Proton dose calculation on scatter- corrected CBCT image: Feasibility study for adaptive proton therapy. Medical Physics. 2015;42(8):4449–59. Keywords: Adaptive radiotherapy, CBCT scatter correction Digital Poster 5215 Predict the three-dimensional spatial contour of target volume based on the body surface movement characteristics of patients with thoracic tumors Yimei Liu 1 , Ruotong Chen 1 , Zixuan Leng 2 , Jie Dong 3 , Meining Chen 1 , Jun Zhang 1 , Guangyu Wang 1 , Yinghui Li 4 , Xiaowu Deng 1 , Yinglin Peng 1 1 Department of Radiation Oncology, State Key Laboratory of Oncology in South China, Guangdong Key Laboratory of Nasopharyngeal Carcinoma Diagnosis and Therapy, Guangdong Provincial Clinical Research Center for Cancer, Sun Yat-sen University Cancer Center, Guangzhou, China. 2 Department of Radiation Oncology, State Key Laboratory of Oncology
Made with FlippingBook - Share PDF online