Quantitative Analysis of immunohistological stainings (segmentation)

fiji
imagej

#1

Hey there!

I’m pretty new in image analysis, and for my project i have to quantify the amount of Mucin (stained blue) and compare it to the amount of inflamed epithelial tissue (stained in pink).
I already have a pretty good method for quantifying the Mucin by colour deconvolution, but the problem lies with the pink quantification: I need only the epithelial tissue, not the Muskularis that surrounds it. (see example image)

I already tried using the weka trainable segmentation, but unfortunately the images (even when i downsize them to 10% of their original size) are too big, therefore fiji can’t keep up. Is there any other way to segment those different tissues?

Thank you in advance!
Vorzeige%20Forum


#2

Good day,

did you try to convert your RGB-images to Lab.

The classification may be a bit easier in this colour space.

Regards

Herbie


#3

Hey, thanks for taking your time to answer my question!

What exactly do you mean? Converting the images and then using weka? or something else entirely?

Thank you!


#4

Well,

the image you’ve posted appears not to be an original raw image and I’ve no experience with the tissue and the staining.

My impression was, that the suggested colour conversion may make the segmentation a bit easier, although I must admit that Muscularis and Epithelial appear quite similar in the sample image. Just try with the original image and maybe you will find a contrast. If not, you have to look for the texture differences which is more involved. WEKA may be able to distinguish the two regions but surely not from the provided sample image. Try with the best spatial resolution possible.

Good luck

Herbie