Adding a parameter to control for number of nuclei in each myotube

[KYazgan1]

Hi, I was wondering if the developers could add another parameter in the settings that could indicate the minimum number of nuclei in a myotube to be counted as positive.

For example, I want to be able to count all positive nuclei in a myotube that contains at least another 2 fused nuclei. If the myotube contains less than 3 fused nuclei, I don't want those nuclei to be deemed positive. I believe that this in general how the field wants fusion index to be measured. Could the developers add this feature?

Thanks very much!

[WeisLeDocto]

Hello,

I believe that this in general how the field wants fusion index to be measured

I understand your concern, especially for situations where the fusion index is low and many myoblasts are still unfused.

Could the developers add this feature?

Depends on what kind of implementation you have in mind.
If it is implementing a simple filter that checks for each contiguous positive myotube area how many nuclei it contains, and makes the nuclei in too poorly populated areas negative, then yes that's possible.
Anything more complex will likely be too much to implement (e.g. identifying two overlapping myotubes as separate objects, implement a more robust myotube detection than the basic thresholding we perform, etc.)

If you can find a reliable segmentation tool that can better detect myotubes, you can let me know and I might consider using it if relevant. A quick Google search sent me to The Myotube Analyzer, that uses a skeleton-based approach on top of a classic thresholding. I might give it a try, but it'll never be as effective as a well-trained AI model (which I don't think currently exists for this task).

Not directly related but interesting IMO, I also came across this preprint, where they use CellPose for nuclei segmentation, and a home-made model for discriminating myoblasts located above or below a myotube (that should therefore be excluded). It's based on the fact that a positive nucleus should leave a "hole" in the myotube signal. It is indeed often the case, but is that always true ? I'd need to further check on some images.

[KYazgan1]

Hey! Yes, the simple filter is the one I had in mind. Basically I'd like to not count any nuclei in any contiguous positive myotube area that includes less than 3 nuclei.

I've been optimizing the program for some of my images and the segmentation is working well so far, of course its no watershed, but I've noticed it's detecting two nuclei that are really close together if not slightly overlapping. But those sources look like an interesting way to improve it!

Thanks!

[WeisLeDocto]

It's now implemented, there's a new Minimum nuclei count setting in the settings window.
When processing an image, the positive nuclei located in fibers that do not contain at least that number of positive nuclei are all re-labeled as negative.
Works fine on my side.
There should be no issue loading projects from the previous version to this one, and vice-versa.

Warning:
Since I also fixed a faulty step in fiber segmentation, the results obtained on a same image with versions 1.0.11 and 1.0.12 may vary, even when setting the minimum nuclei count to 0.

You can update by running: python3 -m pip install --upgrade myofinder