[Insight-users] Metrics for assessment of segmentation result

[Luis Ibanez]

Hi Jim,

The typical reference for segmentation evaluation is "Valmet":
http://midag.cs.unc.edu/pubs/papers/MICCAI01-gerig-valmet.pdf

You can get a binary download of the Valmet tool for free,
but unfortunately its source code is not available.        :-(


On the other hand,
All of the metrics described in the Valmet paper
are now implemented in ITK   :-)


You may want to look at:


A) Overlap metric:
http://www.itk.org/Insight/Doxygen/html/classitk_1_1KappaStatisticImageToImageMetric.html

    Note that when the images are aligned, this could be reimplemented
    to run a lot faster in an ImageCalculator.


B) Hausdorff Distance:
   Directed:
http://www.itk.org/Insight/Doxygen/html/classitk_1_1DirectedHausdorffDistanceImageFilter.html
   Symmetric:
http://www.itk.org/Insight/Doxygen/html/classitk_1_1HausdorffDistanceImageFilter.html


C) Mean Surface Distance:
   Directed:
http://www.itk.org/Insight/Doxygen/html/classitk_1_1ContourDirectedMeanDistanceImageFilter.html
   Symmetric:
http://www.itk.org/Insight/Doxygen/html/classitk_1_1ContourMeanDistanceImageFilter.html



Also, if you need to build concensus from multiple manual segmentations 
you may be interested in the STAPLE method:
http://www.itk.org/Insight/Doxygen/html/classitk_1_1STAPLEImageFilter.html




   Enjoy Open Source !



      Luis



------------------------------
abatzis at ceid.upatras.gr wrote:
> 
> 
> Dear Luis/All,
> 
> i want to assess the result of the 3d segmentation performed by use of the
> level sets, compared to a manual segmentation result. Is there any metrics
> implemented for such purpose or do i have to make my own piece of code?
> 
> Thanks in advance
> Best Regards
> Jim
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users
>