Uncertainties in the use of synthetic CT (sCT) for daily online adaptive radiotherapy for lung cancer

Sammanfattning: Background and purpose: Cone Beam Computed Tomography (CBCT) based online adaptive radiotherapy (oART) has been clinically implemented for several anatomical sites worldwide. However, random and abrupt anatomical changes with large density alterations can occur in lung patients, leading to uncertainties with synthetic computed tomography (sCT) and dose calculation. The anatomical changes could be the resolution/appearance of atelectasis, tumor shrinkage, tumor positional deviation, pleural effusion, pneumonitis, etc. This study aimed to quantitatively investigate the uncertainty of dose calculation on sCT in the Ethos (Varian Medical Systems, Palo Alto, CA) oART-workflow and the possibility of using direct dose calculation on CBCT to minimize the dosimetric uncertainties for lung cancer patients with a broad range of anatomical changes. As well as qualitatively observe the generated sCT images. Methods: Twenty lung cancer patients with different anatomical changes, resulting in re-planning during their courses of treatment, were included (treated between 15 January 2018 and 15 December 2022). An oART workflow was simulated for each patient in an Ethos emulator running in batch mode (Varian Medical Systems). During the simulated oART sessions, the original planning CT (pCT) was deformed to the re-planning CT (rCT), and also deformed to the first CBCT post-re-planning, used as input for the anatomy-of-the-day, generating the sCTrCT and sCTCBCT, respectively. The rCT treatment plan was re-calculated on the generated sCTrCT, sCTCBCT, and directly on each patient's CBCT (dCBCT). Resulting dose distributions were compared with the rCT using relevant dose volume histogram (DVH) parameters, where the dosimetric agreement was evaluated using Wilcoxon signed rank testing. Results: 12 out of 20 generated sCTrCT and 14 out of 20 sCTCBCT images were assessed to have an anatomical disagreement with rCT due to anatomical changes that could not be corrected for by deformation in the Ethos oART workflow. The DVH differences between rCT-sCTrCT (in 13 out of 20 patients) and rCT-sCTCBCT (in 11 out of 16 patients, four patients were excluded due to large anatomical disagreement between CBCT and rCT) were acceptable (within 2.0%). The rest of the patients resulted in higher (over 2.0%) dose differences for any dosimetry metric investigated due to incorrect image deformations and erroneously generated sCTs because of anatomical changes. The DVH dosimetry metric differences in rCT-sCTrCT (median, [min; max]) for the target structures were 0.3 [-2.0; 4.0]%, 0.3 [-0.2; 3.4]%, and -0.4 [-7.8; 0.9]% for GTV D99%, PTV D98% and PTV Dmax, respectively. In contrast, the corresponding differences for organs at risk (OAR) were 0.0 [-0.2; 0.1]%, 0.0 [-0.3; 0.2]%, 0.0 [-0.7; 1.0]%, 0.0 [-1.1; 0.5]% and -0.4 [-7.2; 2.0]% for mean lung dose (MLD), esophagus mean dose (Dmean), heart V25Gy, spinal cord maximum dose (Dmax) and body Dmax, respectively. The Wilcoxon signed rank presented statistically significant differences between rCT-sCTCBCT and rCT-sCTrCT, where DVH differences for rCT-sCTCBCT were -1.3 [-9.0; 1.4]% and -1.0 [-8.1; 1.4]% in comparison to rCT-sCTrCT with -0.4 [-7.8; 0.9]% and -0.4 [-7.2; 2.0]% for PTV Dmax and body Dmax, respectively. The corresponding DVH differences between dCBCT and rCT were 1.1 [-1.3; 5.1]% for PTV D98%, and OAR, the differences were -1.9 [-6.8; -0.1]%, -7.2 [-15.2; 1.6]%, -1.7 [-27.1; 7.6]%, -0.7 [-7.4; 4.7]%, and -2.2 [-5.7; 4.3]% for MLD, esophagus Dmean, heart V25Gy, spinal cord Dmax and body Dmax, respectively. Conclusions: Anatomical disagreement between sCTs and rCT images for lung cancer patients with anatomical changes could cause dose differences of over 2.0%. However, this is not always the case, and therefore this type of patient cohort with anatomical changes undergoing oART must be carefully evaluated on a patient individual level. This finding demonstrates the need for patient and fraction-specific quality assessment, which is unavailable in the current version of Ethos treatment system. The future solution could potentially be direct dose calculation on CBCT. However, this requires further investigation since no conclusion can be drawn from this study due to rCT and CBCT geometrical differences. The findings in this study suggest that using the oART workflow with the Ethos treatment system for lung cancer patients with anatomical changes is currently not clinically appropriate due to the uncertainties with the resulting sCT used for the oART workflow.