Chris Pudney writes:
> The reviewers are probably right - you should easily be able to
> outperform thresholding.  One of the problems with thresholding is that
> a good threshold at the top of the stack is no good deeper in the stack
> as the intensity often falls-off with depth along the optical axis.  Is
> anyone aware of published methods for correcting this?  Then
> thresholding might do better.

Chris--

Anders Liljeborg's paper (see below) describes an approach to that problem--

Liljeborg A.  Czader M.  Porwit A.  A method to compensate for light attenuation
with depth in three-dimensional DNA image cytometry using a confocal scanning
laser microscope.    Journal of Microscopy.  177 ( Pt 2):108-14, 1995

--One issue that may be worth noting in this discussion is that some problems of
segmentation arise from the lack of infinitely thin resolution in the z-axis
with a confocal microscope: thick structures appear brighter than thin
structures, assuming equal concentrations of fluorophore in both.  Deconvolution
should help with that.

Martin Wessendorf