ESTRO 2024 - Abstract Book

S3760

Physics - Image acquisition and processing

ESTRO 2024

The results of this study show that the predicted 3DCTs are of sufficient quality for dose calculation. The proposed method may be considered for online or offline proton therapy treatment adaptation.

Keywords: Dosimetric evaluation, synthetic 3DCT, CNN

References:

[1] Loÿen E, Dasnoy D, Macq B, Patient-specific three-dimensional image reconstruction from a single X-ray projection using a convolutional neural network for on-line radiotherapy applications, Physics and Imaging in Radiation Oncology, 2023, https://doi.org/10.1016/j.phro.2023.100444.

[2] Wuyckens S, Dasnoy D, Janssens G, Hamaide V, Huet M, Loÿen E, et al., OpenTPS – Open-source treatment planning system for research in proton therapy, arXiv, 2023, https://doi.org/10.48550/arXiv.2303.00365.

400

Mini-Oral

Constructing high-quality enhanced 4D-MRI with personalized modeling for liver cancer radiotherapy

Xinyuan Chen, Ke Wang, Bo Chen, Ying Cao, Lijing Zuo, Kaixuan Zhang, Kuo Men, Jianrong Dai

National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Department of Radiation Oncology, Beijing, China

Purpose/Objective:

For liver cancer radiotherapy, four-dimensional (4D) imaging has been found to be an important simulated technology. Compared to CT simulation, magnetic resonance simulation has the advantages of nonionizing radiation and superior soft-tissue contrast. However, MRI acquisition time and image quality are two mutually conflicting indexes, and this conflict may render four-dimensional MRI (4D-MRI) difficult. The application and commercialization of 4D-MRI using the existing MRI sequences are currently limited by the long reconstruction time or poor image quality. Herein, we developed a high-quality personalized enhanced 4D-MRI method to delineate tumor/OARs in different respiratory phases. To overcome the aforementioned deficiencies of limiting its application to the clinic, we first established low-quality 4D-MR image scanning strategies based on the enhanced LAVA sequence. Then, we adopted deep-learning methods to generate high-quality 4D-MR images and utilized a personalized strategy for dealing with the complexity of MRI imaging and the diversity of cancer patients.

Material/Methods: