Comparison between margin-growing algorithms in radiotherapy software environments

Margin-growing algorithms are commonly used tools that are available within virtual simulation and treatment planning software. We report on the accuracy of the margin-growing algorithms available in six commercially available radiotherapy software environments. A phantom containing two differently sized spheres and two rods (one level and one inclined) was constructed and scanned by CT with 1.25 mm, 2.5 mm, 3.75 mm and 5 mm slice thicknesses. The objects were outlined on a GE Advantage Simulator, and the outlined volumes recorded. Images and structures were transferred to MasterPlan, Xio, Pinnacle, Eclipse and Prosoma, where imported volumes were recorded. The contours on each system were grown isotropically by 10 mm, 20 mm and 30 mm, and volumes for each grown contour were recorded. Transfer of structure sets created in GE Advantage Simulator to the other software environments showed that the reported volumes of the four structures differ on each system. Results showed no correlation between volume accuracy and slice thickness. In general, margin growth of up to 30 mm for the rods and spheres is shown to be consistent between systems to within 1.33 mm for all slice thicknesses. Slice thickness did not appear to influence the accuracy of margin growth. Although this work highlights apparent differences in the reported volumes grown from the same original structure sets, the significance of this aspect of the planning process needs to weighed against reported intra- and inter-clinician variability in contour definition. It is not unreasonable, however, to expect that software packages should at least be consistent in volume information provided to the user.