Biovoxxel introduces new cell borders in segmented images

Hey all,
I am trying to analyze some tissues for cell shape and neighbor numbers. For this I created a segmented image with Tissue Analyzer and start with the nice segmented image.
However, whenever I run neighbor analysis or shape descriptor maps, I found new cell borders introduced, especially in stretched cells.
I tried to run the tools without doing “run watershed first” but this is not working at all, it seems that the tool does not consider this image being already segmented.
Here is an example: One introduced cell border can be found at the lower right.
Voronoi_handCorrection.tif-1.tif (36.6 KB) handCorrection.tif (105.0 KB)
Is there any chance to circumvent the problem?

All the best

Hi Anja,

the problem in your image is that the particles cannot be detected as separate. If you zoom in and check the borders, most pixels of neighbouring particles touch each other at the pixel corners. In ImageJ this cannot be detected as separate objects. The same would happen, if you e.g. run the Analyze Particles function.

Solution: Run an 1-pixel erosion (→Process→Binary→Erode). Potentially you will need to use a dilation instead depending on the binary status of you image (foreground - background color white or black and the setting “Black Background” under →Process→Binary→Options…

grafik original

grafik after erosion / dilation
This separates the particles and you can run the Neighbor Analysis without the watershed function. The latter often introduces overfragmentation, therefore is often not suitable in images like yours

grafik analysis

If you finally don’t like the separated appearance, you can run →Process→Filters→Maximum… on the final output image to blow up the objects again
grafik after Maximum… (Filter)

Hope this clarifies the issue and solves the problem

Wow, thanks for the fast help. I will try this :slight_smile:

1 Like