[Insight-users] Metric insight analysis

[abatzis at ceid.upatras.gr]

  Dear Luis and itk collegues,

i've used two itk algorithms so as to perform 3d segmentation of carotids
and so as to assess the results obtained, i used kappa statistic, match
cardinality and hausdorff distance metrics. The ground-truth standard i
used so as to compare my results, on the basis of the aforementioned
metrics is a manual delineation of the datasets provided, by use of the
itk-snap tool. What i'd like to be explained of is how these metrics work,
that is how  the comparison is performed. Do these metrics operate on 2d
space by comparing the automated and manual results slice by slice and
then providing an average of the 2d results or some kind of interpolation
is performed to the manual segmentation and the comparison is performed on
the 3d space? Furthermore the unit results of the hausdorff metric is in
pixels or voxels?

Thanks in advance
 Jim Abatzis