Re: Trichrome Stain Analysis

[Francis Burton]

Hi,

I previously posted to recommend k-means clustering for segmentation of
Masson's Trichrome stained histology images. Having played around with this
some more, including with a sample image that Dayana sent me, I have come
to the conclusion that it can work well in some cases but not all, and that
David Webster's advice is the more sound.

The k-means clustering plugin at
http://ij-plugins.sourceforge.net/plugins/clustering/index.html
works with RGB values, when used with colour images, rather than hues (e.g.
in HSV space). As a result, colours which are very distinct to the eye but
close in RGB space (and grey level) are not distinguished, leading to an
ineffective segmentation.

This doesn't detract from the idea of k-means clustering, and no doubt it
works well in other applications. However, is *not* reliable for Masson's
Trichrome segmentation.

I am wondering whether it could be made to work better in this case if the
images were transformed into HSV space before segmentation is performed.

Francis

At 13:32 16/06/09 -0400, Dayana Webster <[hidden email]> wrote:
>After staining with Masson's Trichrome stain, I am analyzing
>immunohistochemistry images for the amount of blue collagen vs. red muscle
>fibers.  I am having trouble finding a plug-in the can help me distinguish
>and count the blue vs. red areas in the images.  Could anyone help me?

[azizia abdullah]

Dear all,

I am doing research on image categorization. One of the challenging tasks
for me is to construct feature representation from an input image. Thus,
I have a simple question pertaining the feature representation. I have constructed a set
of euclidean distances from an input set of n vectors. The distances are
computed between the set of n vectors and cluster centers or centroids.
After that, I want to use all distances as one of features for indexing.
My question is what is the best way to represent the euclidean distances
for indexing? I know that one of the commonly used approaches is histogram
i.e. histogram of euclidean distances. Does anybody knows what is the best
approach rather than histogram?

kind regards
azizisya.