Help with segmentation of muscle section

Dear all,

I am trying to measure the cross-sectional area of muscle cryosections. I have tried to do segmentation of the image on ImageJ without success.

Any tools you would recommend for this?

Thanks

Carlos

Hi and welcome to the forum!

It would be great if you could specify with that you mean specifically by cross-sectional area.
Are these these smaller sections like so?


Maybe you could point that out or outline that in the provided images.

Thank you very much and sorry for not providing enough information about the image - new in the forum :slight_smile:

Yes, you are right; these are muscle fibers that I want to measure the area.

The idea is being able to do segmentation like this:

It is a difficult problem I think. Segmenting the entire muscle area seems easy, there is a lot of difference in texture and color to do that with machine learning. Robustly segmenting individual muscle sections is the problem. In the top image there is a lot of variability in how these muscle sections are separated. Sometime there is a large bright separation and sometimes there is a darker separation.

The best idea I have is to look into ilastik: https://www.ilastik.org/
You can try to separate the different sections using the carving option in ilastik: https://www.ilastik.org/documentation/carving/carving
They show it in EM but your problem looks similar to this.

1 Like

Hi,

Following @schmiedc suggestion, I suggest to look also at Ilastik boundary based segmentation workflow.

You’ll need first to train a pixel classifier to detect the “borders between cells” both bright and dark ones. If simple Pixel classifier does not work well enough, try the Autocontext pixel classifier workflow .
Then as a second step, you’ll train the boundary-based multicut classifier.

best
Ofra

1 Like

Hello,

out of the box, the result output by CellPose doesn’t look too bad:

(I just changed the average cell diameter to 20 for your embedded image).

There is a GUI on their GitHub page if this is an avenue you want to explore.

Cheers,
Egor

4 Likes