3007 - Robust Bias Dose Optimization for Abdominal Reirradiation within an MR-Guided Online Adaptive Radiotherapy Workflow

Purpose/Objective(s): Reirradiation of abdominal lesions with local control intent is particularly challenging due to concerns regarding toxicity from cumulative doses and significant interfractional anatomical variation of adjacent organs at risk (OARs). This study evaluates the feasibility of MR-guided online adaptive radiotherapy incorporating robust bias dose optimization in the reirradiation setting.

Materials/Methods: Retrospective data from two cases (liver and pancreas) treated with a hypofractionated 5-fraction regimen on an Elekta Unity MR-Linac were analyzed. A workflow was implemented in MIM (MIM Software, Cleveland, OH) to deformably map doses from the prior treatment planning image (PPI) to the reference planning MRI (RP-MRI) and daily planning MRIs (DP-MRI) using a feature similarity metric deformable image registration (DIR) algorithm, with additional targeted local OAR corrections based on a contour-based DIR algorithm, including estimation of voxel-wise dose mapping uncertainty bounds. Mapped doses were converted to equivalent dose in 2 Gy fractions (EQD2) and subsequently to a physical bias dose corresponding to the retreatment fractionation scheme. The bias dose was transferred to a research version of RayStation (RaySearch Laboratories, Stockholm, Sweden) incorporating magnetic fields and used to optimize the reference plan to meet cumulative EQD2 and de novo physical dose limits. Full reoptimization was performed on each DP-MRI using the corresponding mapped bias dose and the reference plan objectives and constraints as the starting point.

Results: The time required to deformably map doses to the DP-MRIs, estimate their uncertainty, and convert them to a bias dose equivalent to the retreatment fractionation ranged from 3 to 7 minutes, depending on the number of OAR contour-based DIR corrections. Among all OARs considered, the hot spot locations from prior treatment were consistent between the RP-MRI and DP-MRIs and with the PPI, with dose-volume metric differences (e.g., EQD2_0.03cc) averaging 40 cGy (range, 7-87 cGy), well within the estimated metric uncertainty (average, 120 cGy, range 75-155 cGy). Optimization time varied from 2 to 3 minutes, with minimal changes to the plan objective weights required to achieve daily target coverage in the range of 96.5%-99.5%. Cumulative EQD2 limits were met in all adaptive plans, and uncertainty bounds representing best- and worst-case scenarios were informed.

Conclusion: Abdominal reirradiation based on robust bias dose optimization is feasible within MR-guided online adaptive radiotherapy workflows. This approach minimizes the risk of toxicity to adjacent OARs while maximizing target coverage and enhancing patient safety.