Purpose: For structures abutting steep dose gradients, small amounts of geometric uncertainty can significantly affect constraint compliance. In this work, we strive to quantify plan robustness to geometric uncertainty with respect to DVH objectives, report a digestible risk index for plan evaluation, and highlight areas for increased consideration during patient set-up.
Methods: To calculate Geometric Risk Index (GRI) for a DVH objective on a per-voxel basis, we first calculated dosimetric and geometric margins (DM, GM). DM was defined as the difference between the objective threshold and achieved plan value. GM was defined as the distance along the steepest dose gradient (SDG) to reach the objective threshold. Geometric uncertainty (GU) was user defined. Dosimetric uncertainty (DU) was calculated as the change in dose along the SDG at a distance of GU. The ratios of GM/GU and DM/DU were combined in a Euclidean space. The inverse of the magnitude of the subsequent vector was reported as GRI. To assess GRI viability, we first investigated maximum dose constraints. A plug-in was developed within Varianâ€™s Eclipse treatment planning system to test on clinical data.
Results: Preliminary results confirm GRI sensitivity to dose gradients and tight dosimetric or geometric margins. As margins increase relative to the uncertainties, GRI decreases to zero, indicating the plan has high robustness to uncertainty. In contrast, when uncertainty exceeded the margin, GRI exceeded the passing threshold of 1, indicating that a voxel was at risk to violate the selected objective at the current level of geometric uncertainty.
Conclusion: Preliminary tests of the proposed metric show promise for providing an easily interpreted measure of risk which accounts for geometric uncertainty and volumetric dose information. With further development and improved sophistication, this metric could become a valuable tool to assess plan quality and guide focus for patient alignment during image-guided set-up procedures.