Hello!
I compute Dice and Hausdorff Distance, you can see the result. So, I know that Dice similarity ranged from 0 to 1. How Hausdorff distance was ranged? Consequently, how to interpret this result. Any explanation will be helpful. Thanks!
Hausdorff distance is millimeters.
Here’s the wikipedia page for the Hausdorff distance that explains the concept with a helpful figure
Here is the doxygen page from the Plastimatch Hausdorff algorithm with equations (this is what the module uses, need to scroll down a bit)
http://plastimatch.org/doxygen/classHausdorff__distance.html
By the way this last link is referenced from the module documentation
https://www.slicer.org/wiki/Documentation/Nightly/Modules/SegmentComparison
It is generally a good idea to look at the documentation, as many times you will find useful information there.
Thank!
on a related note…does anyone have any information on whether the plastimatch Hausdorff or DICE calculations have been validated against any other viable option?
Yes, we validated it against analytic solids prior to using it to score the MICCAI 2015 H&N Grand Challenge. That means comparing translated and resized cubes, spheres, and cylinders, and comparing against the analytic solution.
Thanks Greg - are there any publications or links to details available anywhere?
We’d like to use slicerRT to assess AI contouring quality in a research project.
Hi John. I checked, but that information was not included in the paper. I guess you could cite as “Personal communication”. Sorry.
Thanks for checking Greg. Much obliged.