Room: Exhibit Hall
Purpose: To develop an open source toolkit to easily view, evaluate and compare dose volume histogram (DVH) data as well as perform independent DVH calculations. The toolkit was used to evaluate various methods of interpolation used in independent DVH calculation for small volume structures and high dose gradients and to compare with existing treatment planning system DVHs.
Methods: Using the Python programming language, a toolkit "dicompyler-core" was developed to allow users to import DVH data from treatment planning systems via DICOM or via raw histogram data. It was designed to allow easy access to the underlying data with the ability to transform DVHs between relative and absolute units and between cumulative and differential modes, as well as functions to plot and compare DVHs from different sources. The DVH calculation algorithm from the dicompyler project was enhanced by super-sampling the resolution during voxelization of the structure and dose grid in the in-plane direction. The super-sampling was restricted to the boundaries of the structure to not significantly increase calculation time. Additionally, in the axial direction, structure and dose grid planes were interpolated to a finer sample spacing than the original spacing.
Results: dicompyler-core was able to successfully compare the differences between the independently calculated and treatment plan DVHs. For the interpolation methods, the in-plane direction super-sampling showed significant improvement compared to sampling with the standard dose grid resolution. In the axial direction, the sample spacing made a significant difference only if the structure contours changed significantly between planes. Comparing with existing treatment planning systems showed agreement within 2.5% for volume calculation as well as standard metrics like D0.03cc.
Conclusion: Interpolation resolution and direction are important factors when calculating DVHs for small volume structures and high dose gradients. Evaluation using a toolkit like dicompyler-core facilitates quick comparison to determine optimal interpolation parameters.