ESTRO 2021 Abstract Book

S1370

ESTRO 2021

Truebeam Linac. Precision, accuracy and reproducibility of the treatment plan is of the outmost importance for SRS especially for small target volume. In this work, we perform a case study to study the impact of MLC and gantry errors on the dose coverage of the CTV in a single fraction brain metastasis. Materials and Methods Trajectory log files for 2 single fractionated SRS plans (1 single metastasis with 20 Gy, 1 multiple metastasis with 25 Gy) were extracted from Varian TrueBeam. The verification of the dose distribution for both plans was verified prior to treatment using SRS MapCheck with γ passing rate (%GP) of 95% and above. An in-house script was developed to decode the log files created during treatment. The errors in gantry and MLC positions were compiled for the two cases. Three different modified plans were created incorporating 1) actual gantry, 2) actual MLC and 3) actual MLC and gantry positions. These plans were used to recalculate a new dose distribution in the TPS and were compared with the original treatment dose. The values of the gross tumour volumes (GTV) and planning target volumes (PTV) for the dose levels of D99%, D95% and D90% were obtained from the recalculated dose volume histograms (DVH) and evaluated against the original plans to assess any differences. Results The gantry and MLC errors were found to be within ± 0.1 deg / 0.1mm tolerance from the log file. The target volumes shows a loss in dose coverage for the two cases, where the GTV D95 decreases by about 1 Gy. Gantry errors has a larger impact on target coverage than MLC errors for the two cases.

Conclusion The target dose coverage in small target volume is sensitive to MLC and gantry error and the planning process must take into account the possible loss of dose coverage from delivery error. Further study is under way to determine if the sensitivity is a function of plan complexity and use of jaw tracking.

Digital Poster: Imaging acquisition and processing

PO-1651 The dosimetric impact of deep learning-based organs at risk auto-segmentation H. Guo 1,2 , X. Xia 2,1 , Y. Zhong 1,2 , J. Peng 1,2 , W. Hu 1,2 , J. Wang 1,2 , Z. Zhang 1,2 1 Fudan University Shanghai Cancer Center, Department of Radiation Oncology, Shanghai, China; 2 Fudan University Shanghai Medical College, Department of Oncology, Shanghai, China Purpose or Objective To investigate the dosimetric impact of deep learning-based auto-segmentation of organs at risk (OARs) on nasopharyngeal and rectal cancer. Materials and Methods Twenty patients, including ten nasopharyngeal carcinoma (NPC) patients and ten rectal cancer patients, who received radiotherapy in our department were enrolled in this study. Two deep learning-based auto- segmentation systems, including an in-house developed system (FD) and a commercial product (UIH), were used to generate two auto-segmented OARs sets (OAR_FD and OAR_UIH). Treatment plans based on auto- segmented OARs and following our clinical requirements were generated for each patient on each OARs set (Plan_FD and Plan_UIH). Geometric metrics (Hausdorff distance (HD), mean distance to agreement (MDA), the Dice similarity coefficient (DICE) and the Jaccard index) were calculated for geometric evaluation. The dosimetric impact was evaluated by comparing Plan_FD and Plan_UIH to original clinically approved plans (Plan_Manual) with dose-volume indices and 3D gamma pass rates. Spearman’s correlation analysis was performed to investigate the correlation between dosimetric deviation and geometric metrics.