Background
The MR-guided adaptive radiotherapy (MRgART) workflow at the 1.5 T Unity MR-Linac relies on synthetic CT (sCT) generated through bulk density assignment. Although sCT-based dose calculations are the standard approach, it is well known that their accuracy can be compromised in lung tumors due to the high dose gradients surrounding the targets. This study investigates clinical cases to determine whether these challenges affect all lung targets or if, for a subset partially located in high-density dose regions, the sCT calculations performed in the Unity clinical workflow are sufficiently accurate, supporting their routine clinical application.
Methods
Forty-eight lung cancer patients undergoing stereotactic body radiotherapy at Unity MR-Linac, were included in this study. Patients were stratified into two groups based on target position: \(\:{\text{G}}_{\text{A}\text{T}\text{M}}\) group including targets attached to thoracic wall or mediastinum, and \(\:{\text{G}}_{\text{I}\text{P}\text{T}}\) (group isolated pulmonary target), including targets entirely within the lung parenchyma and surrounded by air-filled lung tissue. The reference treatment plans ( \(\:{\text{T}\text{P}}_{\text{R}\text{E}\text{F}}\) ) were optimised on the simulation CT using inverse-planning intensity modulated radiation therapy (IMRT) with the Monaco treatment planning system to deliver 50 Gy in 5 fractions. \(\:{\text{T}\text{P}}_{\text{R}\text{E}\text{F}}\) included all contour information needed to generate the sCT through bulk electron density assignment, the standard procedure for Unity MR-Linac. To evaluate the dosimetric accuracy of sCT-based dose calculations, a second plan ( \(\:{\text{T}\text{P}}_{\text{s}\text{C}\text{T}}\) ) was created by recalculating \(\:{\text{T}\text{P}}_{\text{R}\text{E}\text{F}}\) on the sCT derived from the reference CT. The sCT was generated using the MRgART routine, which assigns mean ED values to all contoured structures. Dose-volume histograms were compared between \(\:{\text{T}\text{P}}_{\text{R}\text{E}\text{F}}\) and \(\:{\text{T}\text{P}}_{\text{s}\text{C}\text{T}}\) for targets and organs at risk (OARs), with additional evaluation of tumor control probability, and normal tissue complication probability. Dose distributions were further evaluated using global gamma analysis with 3%/3 mm and 2%/2 mm criteria.
Results
Significant differences (p < 0.05) in key dosimetric parameters ( \(\:{\text{V}}_{50\text{G}\text{y}}\) , \(\:{\text{V}}_{45\text{G}\text{y}}\) , \(\:{\text{D}}_{50\text{\%}}\) ) were observed between \(\:{\text{T}\text{P}}_{\text{R}\text{E}\text{F}}\) and \(\:{\text{T}\text{P}}_{\text{s}\text{C}\text{T}}\) in the \(\:{\text{G}}_{\text{I}\text{P}\text{T}}\) group, with percentage differences reaching up to 3.49%. Conversely, in \(\:{\text{G}}_{\text{A}\text{T}\text{M}}\) , percentage differences were less than 1% and not statistically significant (p > 0.05). For OARs, no significant differences (p > 0.05) were observed in either group, except for the lungs minus the gross tumor volume (lungs-GTV), where percentage differences remained below 1.5%. Radiobiological modelling yielded consistent results, confirming the dosimetric findings. Gamma analysis showed consistent dose distributions, with a global pass rate above 95% for 3%/3 mm criteria and above 90% for 2%/2 mm criteria.
Conclusions
This study demonstrates that sCT-based dose calculations are feasible and reliable for pulmonary targets attached to the thoracic wall or mediastinum, supporting their routine integration into MRgART workflows on the Unity MR-Linac. However, for isolated pulmonary targets, deviations should be considered when implementing sCT-based planning in clinical practice.
