Difficult segmentation usign tight junction stain

Dear community,


I’m trying to segment and measure cells by relying on a tight junction staining. The staining in not optimal and there is quite some background. However, visually, it is well seen.
I’ve been trying to use different bluring and convolutions as well as the “dilate” and “erode” options to obtain a continuous mask around my cells, but this has proven difficult.

Would anybody know some alternative options and/or trick to obtain a contiuous ROI around cells when the staining is not optimal?

many thanks for your help,

Segmentation.tif (1.8 MB)

Hi Philip,
have you tried seeded watershed?

Hi Ilya,
I didn’t know this feature, I will try it out now

Thanks for the suggestion!

I think one of the most fitting tools for your image analysis is MorphoLibJ: https://imagej.net/MorphoLibJ

Maybe the Morphological Segmentation (https://imagej.net/Morphological_Segmentation) or the Interactive Marker Controlled (https://imagej.net/Interactive_Marker-controlled_Watershed) would work for you right away.

Otherwise I would invest in fitting Image Processing.
Aim is to get your structures as complete as possible.
I would go for a Gaussian Filter and then Background subtraction.

Maybe the MorphoLibJ toolbox has other filters available to get a good edge detection.


Dear Christopher,
thanks for the suggestion, the Interactive Marker Controlled Watershed worked like a charm for a majority of my dataset.


1 Like