ProKnow as Independent "Gold Standard"
Every TPS has its own technology (i.e. software algorithm) to take radiotherapy objects such as dose and structures and, from them, estimate dose volume histogram (DVH) and other important stats. Some DVH algorithms are more accurate than others, and unfortunately they are highly variable [refs. 1, 2].
We have designed the ProKnow DVH engine to be an industry standard in terms of accuracy and quality. We have tested it using a rigorous testing strategy and benchmark datasets which has been recently published [ref. 2]. For a nice background, please see this presentation: DVH-revisited.pdf.
Our DVH and metric calculation engine is used for all patient datasets, regardless of the TPS in which the dose was calculated. This is good, because it drives out variability and ensures the highest accuracy. However, you may notice some differences compared to your TPS results (differences which are usually minor, but can sometimes be significant). This is normal, and is due to how DVH calculations are implemented by different systems.
Three of the most important aspects of DVH calculation (described in the document link, above), are (1) how is dose super-sampling implemented for small or complex volumes, and/or coarse dose grids, (2) how does the system handle volume 'end-capping' at superior and inferior borders, and (3) what dose bin resolution (Gy) is used to discretize the each voxel's dose?
Regarding dose bin resolution, the dose bin width will be calculated dynamically, per structure, to ensure that there are at least 1000 (and up to 10,000) bins along the dose axis for each structure, i.e. from zero dose to the structure's max dose. All of these parameters help ensure a smooth, high-resolution, and accurate DVH curve and extracted points.
In terms of super-sampling to get fine dose voxels per structure, ProKnow will do enough super-sampling to ensure at least 10,000 volume elements ("voxels") per structure, no matter how small. Sometimes this means super-sampling a dose resolution < 0.1 mm! Also, super-sampling will be used for any structure with a volume < 200 cc and/or for dose grid resolution < 3 mm.
To ensure min and max dose are captured accurately per structure, all contoured points (i.e., surface points) will have a point dose interpolated at their exact 3D coordinate. If that point dose is lower than the lowest sampled dose voxel inside the structure, or higher than the highest, the min or max dose for that structure will be updated accordingly.
Finally, on the topic of end-capping, ProKnow will ensure that the structure's inferior and superior border(s) will be extended halfway to the next dose grid slice, but not to exceed 1.0 mm. One particular TPS (Eclipse, from Varian) tends to underestimate structure volumes in these directions, leading to smaller volumes and less capture of steep dose gradients .
The dose bin resolution, super-sampling, min and max dose refinement, and end-capping rules specified above are improvements on the proven method published in literature. For more detail, please refer to the published article by Nelms et al., "Methods, software and datasets to verify DVH calculations against analytical values: Twenty Years Late(r)." 
 Ebert MV, et al. “Comparison of DVH data from multiple radiotherapy treatment planning systems,” Phys Med Biol. 2010 May; 55(11).
 Nelms BE, Stambaugh C, Hunt D, Tonner B, Zhang G, and Feygelman V. "Methods, software and datasets to verify DVH calculations against analytical values: Twenty Years Late(r)," Med Phys. 2015 Aug; 42(8).
 " DVH Revisited: Everything you (probably) already know and maybe some things you don’t (but should)," presented by Ben Nelms as part of 2017 AAMD webinar series for National Dosimetrist's Week.