Author: Hailun Pan, Yingli Yang, Jie Zhang, Yibin Zhang 👨🔬
Affiliation: Department of Radiation Oncology,Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Department of Radiology, Ruijin Hospital, Department of Radiation Oncology, Ruijin Hospital, , Shanghai Jiaotong University School Of Medicine, Shanghai United imaging Healthcare Advanced Technology Research Institute 🌍
Purpose: Accurate patient positioning is critical in radiotherapy (RT) to ensure effective treatment delivery and minimize harming surrounding healthy tissues. Vertebral misalignment during RT setup has been associated with RT mistakes, as reported in previous studies. This study aims to develop and validate an innovative automated tool for efficient patient initial setup. The proposed method ensures robust alignment, maintaining high accuracy even with large initial offsets, while mitigating the risks associated with vertebral misalignment.
Methods: The registration workflow combines the nnU-Net framework with a four-step post-processing pipeline for vertebral identification and localization. The model was trained on 1,053 CT scans, encompassing public datasets and clinical cases, achieving identification rates of 97.99% on public datasets and 99.76% on clinical datasets, with mean localization errors of 1.64 mm and 1.74 mm, respectively. A rigid registration approach based on Singular Value Decomposition (SVD) was employed for initial alignment, followed by refinement using the Iterative Closest Point (ICP) algorithm. Evaluation was conducted on 43 paired planning CT and setup CBCT datasets, where artificial offsets were introduced to simulate clinical positioning errors.
Results: The proposed method achieved mean rotation errors of 1.09° (superior-inferior axis), 2.58° (left-right axis), and 0.09° (anterior-posterior axis), and mean shift errors of 3.37 mm, 3.06 mm, and 2.43 mm along the respective axes. The error values were significantly smaller than vertebral dimensions, demonstrating the method’s robustness and clinical applicability as an initial setup tool to avoid the need for patient tattoos. An experienced clinician reviewed the results, confirming their reliability across diverse scenarios.
Conclusion: This automated tool provides a reliable and efficient solution for RT initial setup. By effectively correcting large misalignments. It holds the potential to serve as a useful tool to improve and simplify RT treatment workflow.