Lumen detection in qupath

I have fluorescence images for spheroids structures and I am interested in measuring the lumen area and the whole spheroid area. I tried to do so using image J but wasn’t able to get the right thresholding to detect lumens. Now, I moved to use the positive cell detection tool in qupath which looked a bit promising in terms of lumen detection, but now I am facing 2 problems:

1 - Lumen and spheroid detection wan’t accurate, I tried to change the parameters but didn’t have much effect (image attached).
fluorescence image.tif (333.4 KB)
spheroid-gp135-dapi_z12c3 4.tif (6.0 MB)

2- Each image has different analysis parameter and having hundreds of images makes the analysis more complicated time consuming

Since I am pretty new to qupath is there a better tool in qupath that I can use for lumen detection?

Thank you for your help!


What version of QuPath are you using, and if you are using M9/M10 (current test versions), have you tried the pixel classifier?

As it is, I’m seeing some decently large background effects, so it could be that analysis will be difficult until the image can be taken more cleanly.

Odd border at the top, center of field of view noticeably brighter than the edges.

Automated methods work best when the images taken are “fair” to the analysis being conducted.

Thank you @Research_Associate

I am using QuPath m9 and I tried using the simple thresholder in pixel classifier but it is not any better in determining the lumen borders. I preferred “positive cell detection” tool because it gives the size of the lumen and its corresponding spheroid.
I will try your suggestion regarding reducing the background, is it possible to do that in QuPath or I have to do it in another software?

Either FIJI or retaking the images under different conditions (the latter being the better, but usually more annoying option).

I would expect the pixel classifier would fit your desired result better than the simple thresholder since you don’t seem to have a marker for the lumen.

The positive cell detection will not give the size of the corresponding spheroid, as cell expansion is blind. It is not influenced by the any information external to the nucleus, unless one cell expansion would run into another cell expansion.

As you can see in your own image, the cell border traces the outline of the nucleus, not the spheroid. To test this, you can adjust the Cell Expansion value during Cell detection.