Purpose: To evaluate the impact of inter-fractional variability on the quantification of the focal liver reaction (FLR) threshold dose by establishing an auto-delineation approach of the hypo-intense region in the post-SBRT MRI scans.
Methods: Primovistáµ€á´¹ contrast enhanced MRI scans were obtained 8-12 weeks post-SBRT (50 Gy in 5 fractions, V95%) for five patients. These scans permit direct visualization of the hypo-intensity which corresponds to the FLR in the region of dose deposition. Custom software was developed to auto-delineate these hypo-intense regions, with the inputs consisting of post-SBRT MRI, planned dose distribution, liver and PTV structures. This allows for the isolation of liver volume, where each voxel contains the intensity, dose, and spatial location information. By correlating dose and intensity of voxels within the liver, the hypo-intense region was delineated using intensity and dose thresholding. The dose distribution within the hypo-intense region was assessed for the â€˜plannedâ€™ versus â€˜deliveredâ€™ dose. The delivered dose distribution was estimated using accumulated dose approach, where Cone Beam CTs for each fraction were registered to the planning CT (p-CT). Dose was recalculated on p-CT post registration, and finally dose from each fraction was summed thus obtaining delivered dose distribution. The mean Biologically Effective Dose (BED) was quantified for the hypo-intense region, and Wilcoxon-Signed Rank test was performed.
Results: The median (range) BED average was 260 (183-323) Gy for planned dose distribution, and 251 (139-301) Gy for the delivered dose (p=0.04). The planned dose distribution tends to overestimate the dose delivered to the hypo-intense region, with the highest over-estimation being observed at 109-121% dose levels.
Conclusion: This study confirms the feasibility of an automated and consistent delineation of the hypo-intensity which allows for the quantification of the threshold dose of FLR. Specifically, an overestimation of threshold dose was observed when inter-fractional variability is not taken into account.