Dosimetric validation of ART workflow of MRI in adaptive radiotherapy for head and neck tumors
10.3760/cma.j.cn113030-20240430-00175
- VernacularTitle:基于MR的ART工作流程在头颈部肿瘤放射治疗中的剂量验证
- Author:
Zhenkai LI
1
;
Huadong WANG
;
Yong YIN
;
Zhenjiang LI
Author Information
1. 山东省肿瘤医院放射物理技术科，济南 250117
- Keywords:
Radiotherapy, magnetic resonance guided;
Magnetic resonance linac;
Deep learning;
synthetic CT;
Dose verification
- From:
Chinese Journal of Radiation Oncology
2024;33(12):1112-1118
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To verify the synthetic CT(sCT) images generated by the adaptive radiation therapy (ART) workflow guided by magnetic resonance imaging (MRI), and to evaluate its feasibility for dose calculation in ART for head and neck.Methods:The plan CT (pCT) and T 2WI MRI images before each fraction of treatment were collected from 300 patients with head and neck tumors admitted to Shandong Cancer Hospital from January 1, 2022 to December 1, 2023. MRI was converted into sCT using RegGAN model. Among 300 patients, 240 cases were used for training and validation and 60 for testing. The mean absolute error (MAE), peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM) were used to evaluate the image quality of sCT. The original pCT plan conditions were copied to sCT, and the accuracy of sCT was recalculated and verified. Finally, the radiation treatment plan of sCT was verified, and the γ-passing rate was calculated to further verify the accuracy of sCT dose calculation. Results:In terms of index calculation, the MAE of sCT was 32.50±5.46, 31.39±1.14 for PSNR, and 0.93±0.01 for SSIM, respectively. Slight difference was observed in the comparison of CT values. Regarding dose calculation, the accuracy of sCT dose calculation was high. According to the dose volume histogram (DVH), the difference in dose calculation was slight, and the average dose difference was only 0.16%. sCT exhibited a high degree of coincidence with pCT in the evaluation of target delineation for planning target, clinical target and gross tumor volume. The average Dice similarity coefficients were 0.97, 0.95 and 0.94, respectively. The γ-passing rate reached 100% using 3%/2 mm. The total treatment time of all patients was (155.14±9.41) s.Conclusions:The ART workflow has high work efficiency, and the accuracy of the generated sCT can meet the dose calculation accuracy requirements of the online plan. It can be used as a reference image for MRI-guided accelerator ART of nasopharyngeal carcinoma, and is expected to achieve MR radiotherapy based on nasopharyngeal carcinoma.
