Deformable gel dosimetry II: Experimental validation of DIR-based dose-warping

Abstract

Algorithms exist for the deformation of radiotherapy doses based on patient image sets, though these are sometimes contentious because not all such image calculations are constrained by appropriate physical laws. By use of a deformable dosimetric gel phantom, 'DEFGEL', we demonstrate a full 3D experimental validation of a range of dose deformation algorithms publicly available. Spatial accuracy in low contrast areas was assessed using "ghost" fiducial markers (digitally removed from CT images prior to registration) implanted in the phantom. The accuracy with which the different algorithms deform dose was evaluated by comparing doses measured with the deformable phantom to warped planned doses, via 3D g-analysis. Mean spatial errors ranged from 1.9 mm with a g3D passing ratio of 95.8 % for the original Horn and Schunck algorithm to 3.9 mm with a g3D passing ratio of 39.9 % for the modified demons algorithm.